Provided by: slurm-client_19.05.5-1_amd64 
NAME
scontrol - view or modify Slurm configuration and state.
SYNOPSIS
scontrol [OPTIONS...] [COMMAND...]
DESCRIPTION
scontrol is used to view or modify Slurm configuration including: job, job step, node, partition,
reservation, and overall system configuration. Most of the commands can only be executed by user root or
an Administrator. If an attempt to view or modify configuration information is made by an unauthorized
user, an error message will be printed and the requested action will not occur. If no command is entered
on the execute line, scontrol will operate in an interactive mode and prompt for input. It will continue
prompting for input and executing commands until explicitly terminated. If a command is entered on the
execute line, scontrol will execute that command and terminate. All commands and options are
case-insensitive, although node names, partition names, and reservation names are case-sensitive (node
names "LX" and "lx" are distinct). All commands and options can be abbreviated to the extent that the
specification is unique. A modified Slurm configuration can be written to a file using the scontrol write
config command. The resulting file will be named using the convention "slurm.conf.<datetime>" and located
in the same directory as the original "slurm.conf" file. The directory containing the original slurm.conf
must be writable for this to occur.
OPTIONS
-a, --all
When the show command is used, then display all partitions, their jobs and jobs steps. This causes
information to be displayed about partitions that are configured as hidden and partitions that are
unavailable to user's group.
-d, --details
Causes the show command to provide additional details where available.
--federation
Report jobs from from federation if a member of one.
-F, --future
Report nodes in FUTURE state.
-h, --help
Print a help message describing the usage of scontrol.
--hide Do not display information about hidden partitions, their jobs and job steps. By default, neither
partitions that are configured as hidden nor those partitions unavailable to user's group will be
displayed (i.e. this is the default behavior).
--local
Show only information local to this cluster. Ignore other clusters in the federated if a member of
one. Overrides --federation.
-M, --clusters=<string>
The cluster to issue commands to. Only one cluster name may be specified. Note that the SlurmDBD
must be up for this option to work properly. This option implicitly sets the --local option.
-o, --oneliner
Print information one line per record.
-Q, --quiet
Print no warning or informational messages, only fatal error messages.
--sibling
Show all sibling jobs on a federated cluster. Implies --federation.
-u, --uid=<uid>
Attempt to update a job as user <uid> instead of the invoking user id.
-v, --verbose
Print detailed event logging. Multiple -v's will further increase the verbosity of logging. By
default only errors will be displayed.
-V , --version
Print version information and exit.
COMMANDS
abort Instruct the Slurm controller to terminate immediately and generate a core file. See "man
slurmctld" for information about where the core file will be written.
cancel_reboot <NodeList>
Cancel pending reboots on nodes.
checkpoint CKPT_OP ID
Perform a checkpoint activity on the job step(s) with the specified identification. ID can be
used to identify a specific job (e.g. "<job_id>", which applies to all of its existing steps) or a
specific job step (e.g. "<job_id>.<step_id>"). Acceptable values for CKPT_OP include:
able Test if presently not disabled, report start time if checkpoint in progress
create Create a checkpoint and continue the job or job step
disable Disable future checkpoints
enable Enable future checkpoints
error Report the result for the last checkpoint request, error code and message
restart Restart execution of the previously checkpointed job or job step
requeue Create a checkpoint and requeue the batch job, combines vacate and restart operations
vacate Create a checkpoint and terminate the job or job step
Acceptable values for CKPT_OP include:
MaxWait=<seconds> Maximum time for checkpoint to be written. Default value is 10 seconds.
Valid with create and vacate options only.
ImageDir=<directory_name>
Location of checkpoint file. Valid with create, vacate and restart options
only. This value takes precedent over any --checkpoint-dir value specified at
job submission time.
StickToNodes If set, resume job on the same nodes are previously used. Valid with the
restart option only.
cluster CLUSTER_NAME
The cluster to issue commands to. Only one cluster name may be specified.
create SPECIFICATION
Create a new partition or reservation. See the full list of parameters below. Include the tag
"res" to create a reservation without specifying a reservation name.
completing
Display all jobs in a COMPLETING state along with associated nodes in either a COMPLETING or DOWN
state.
delete SPECIFICATION
Delete the entry with the specified SPECIFICATION. The two SPECIFICATION choices are
PartitionName=<name> and Reservation=<name>. Reservations and partitions should have no associated
jobs at the time of their deletion (modify the jobs first). If the specified partition is in use,
the request is denied.
errnumstr ERRNO
Given a Slurm error number, return a descriptive string.
fsdampeningfactor FACTOR
Set the FairShareDampeningFactor in slurmctld.
help Display a description of scontrol options and commands.
hold job_list
Prevent a pending job from being started (sets it's priority to 0). Use the release command to
permit the job to be scheduled. The job_list argument is a comma separated list of job IDs OR
"jobname=" with the job's name, which will attempt to hold all jobs having that name. Note that
when a job is held by a system administrator using the hold command, only a system administrator
may release the job for execution (also see the uhold command). When the job is held by its owner,
it may also be released by the job's owner. Additionally, attempting to hold a running job will
have not suspend or cancel it. But, it will set the job priority to 0 and update the job reason
field, which would hold the job if it was requeued at a later time.
notify job_id message
Send a message to standard error of the salloc or srun command or batch job associated with the
specified job_id.
pidinfo proc_id
Print the Slurm job id and scheduled termination time corresponding to the supplied process id,
proc_id, on the current node. This will work only with processes on node on which scontrol is
run, and only for those processes spawned by Slurm and their descendants.
listpids [job_id[.step_id]] [NodeName]
Print a listing of the process IDs in a job step (if JOBID.STEPID is provided), or all of the job
steps in a job (if job_id is provided), or all of the job steps in all of the jobs on the local
node (if job_id is not provided or job_id is "*"). This will work only with processes on the node
on which scontrol is run, and only for those processes spawned by Slurm and their descendants.
Note that some Slurm configurations (ProctrackType value of pgid) are unable to identify all
processes associated with a job or job step.
Note that the NodeName option is only really useful when you have multiple slurmd daemons running
on the same host machine. Multiple slurmd daemons on one host are, in general, only used by Slurm
developers.
ping Ping the primary and secondary slurmctld daemon and report if they are responding.
reboot [ASAP] [nextstate=<RESUME|DOWN>] [reason=<reason>] <ALL|NodeList>
Reboot the nodes in the system when they become idle using the RebootProgram as configured in
Slurm's slurm.conf file. Each node will have the "REBOOT" flag added to its node state. After a
node reboots and the slurmd daemon starts up again, the HealthCheckProgram will run once. Then,
the slurmd daemon will register itself with the slurmctld daemon and the "REBOOT" flag will be
cleared. If the node reason is "Reboot ASAP", Slurm will clear the node's "DRAIN" state flag as
well. The "ASAP" option adds the "DRAIN" flag to each node's state, preventing additional jobs
from running on the node so it can be rebooted and returned to service "As Soon As Possible" (i.e.
ASAP). "ASAP" will also set the node reason to "Reboot ASAP" if the "reason" option isn't
specified. If the "nextstate" option is specified as "DOWN", then the node will remain in a down
state after rebooting. If "nextstate" is specified as "RESUME", then the nodes will resume as
normal when the node registers and the node reason will be cleared. Resuming nodes will be
considered as available in backfill future scheduling and won't be replaced by idle nodes in a
reservation. When using the "nextstate" and "reason" options together the reason will be appended
with "reboot issued" when the reboot is issued and "reboot complete" when the node registers with
a "nextstate" of "DOWN". The "reason" option sets each node's reason to a user-defined message.
You must specify either a list of nodes or that ALL nodes are to be rebooted. NOTE: By default,
this command does not prevent additional jobs from being scheduled on any nodes before reboot. To
do this, you can either use the "ASAP" option or explicitly drain the nodes beforehand. You can
alternately create an advanced reservation to prevent additional jobs from being initiated on
nodes to be rebooted. Pending reboots can be cancelled by using "scontrol cancel_reboot <node>"
or setting the node state to "CANCEL_REBOOT". A node will be marked "DOWN" if it doesn't reboot
within ResumeTimeout.
reconfigure
Instruct all Slurm daemons to re-read the configuration file. This command does not restart the
daemons. This mechanism would be used to modify configuration parameters (Epilog, Prolog,
SlurmctldLogFile, SlurmdLogFile, etc.). The Slurm controller (slurmctld) forwards the request all
other daemons (slurmd daemon on each compute node). Running jobs continue execution. Most
configuration parameters can be changed by just running this command, however, Slurm daemons
should be shutdown and restarted if any of these parameters are to be changed: AuthType,
BackupAddr, BackupController, ControlAddr, ControlMach, PluginDir, StateSaveLocation,
SlurmctldPort or SlurmdPort. The slurmctld daemon and all slurmd daemons must be restarted if
nodes are added to or removed from the cluster.
release job_list
Release a previously held job to begin execution. The job_list argument is a comma separated list
of job IDs OR "jobname=" with the job's name, which will attempt to hold all jobs having that
name. Also see hold.
requeue [option] job_list
Requeue a running, suspended or finished Slurm batch job into pending state. The job_list
argument is a comma separated list of job IDs. The command accepts the following option:
Incomplete
Operate only on jobs (or tasks of a job array) which have not completed. Specifically only
jobs in the following states will be requeued: CONFIGURING, RUNNING, STOPPED or SUSPENDED.
requeuehold [option] job_list
Requeue a running, suspended or finished Slurm batch job into pending state, moreover the job is
put in held state (priority zero). The job_list argument is a comma separated list of job IDs. A
held job can be released using scontrol to reset its priority (e.g. "scontrol release <job_id>").
The command accepts the following options:
Incomplete
Operate only on jobs (or tasks of a job array) which have not completed. Specifically only
jobs in the following states will be requeued: CONFIGURING, RUNNING, STOPPED or SUSPENDED.
State=SpecialExit
The "SpecialExit" keyword specifies that the job has to be put in a special state
JOB_SPECIAL_EXIT. The "scontrol show job" command will display the JobState as
SPECIAL_EXIT, while the "squeue" command as SE.
resume job_list
Resume a previously suspended job. The job_list argument is a comma separated list of job IDs.
Also see suspend.
NOTE: A suspended job releases its CPUs for allocation to other jobs. Resuming a previously
suspended job may result in multiple jobs being allocated the same CPUs, which could trigger gang
scheduling with some configurations or severe degradation in performance with other
configurations. Use of the scancel command to send SIGSTOP and SIGCONT signals would stop a job
without releasing its CPUs for allocation to other jobs and would be a preferable mechanism in
many cases. If performing system maintenance you may want to use suspend/resume in the following
way. Before suspending set all nodes to draining or set all partitions to down so that no new jobs
can be scheduled. Then suspend jobs. Once maintenance is done resume jobs then resume nodes and/or
set all partitions back to up. Use with caution.
schedloglevel LEVEL
Enable or disable scheduler logging. LEVEL may be "0", "1", "disable" or "enable". "0" has the
same effect as "disable". "1" has the same effect as "enable". This value is temporary and will
be overwritten when the slurmctld daemon reads the slurm.conf configuration file (e.g. when the
daemon is restarted or scontrol reconfigure is executed) if the SlurmSchedLogLevel parameter is
present.
setdebug LEVEL
Change the debug level of the slurmctld daemon. LEVEL may be an integer value between zero and
nine (using the same values as SlurmctldDebug in the slurm.conf file) or the name of the most
detailed message type to be printed: "quiet", "fatal", "error", "info", "verbose", "debug",
"debug2", "debug3", "debug4", or "debug5". This value is temporary and will be overwritten
whenever the slurmctld daemon reads the slurm.conf configuration file (e.g. when the daemon is
restarted or scontrol reconfigure is executed).
setdebugflags [+|-]FLAG
Add or remove DebugFlags of the slurmctld daemon. See "man slurm.conf" for a list of supported
DebugFlags. NOTE: Changing the value of some DebugFlags will have no effect without restarting
the slurmctld daemon, which would set DebugFlags based upon the contents of the slurm.conf
configuration file.
show ENTITY ID
or
show ENTITY=ID
Display the state of the specified entity with the specified identification. ENTITY may be
aliases, assoc_mgr, bbstat, burstbuffer, config, daemons, dwstat, federation, frontend, job, node,
partition, powercap, reservation, slurmd, step, topology, hostlist, hostlistsorted or hostnames ID
can be used to identify a specific element of the identified entity: job ID, node name, partition
name, reservation name, or job step ID for job, node, partition, or step respectively. For an
ENTITY of bbstat or dwstat (they are equivalent) optional arguments are the options of the local
status command. The status commands will be executed by the slurmctld daemon and its response
returned to the user. For an ENTITY of topology, the ID may be a node or switch name. If one
node name is specified, all switches connected to that node (and their parent switches) will be
shown. If more than one node name is specified, only switches that connect to all named nodes
will be shown. aliases will return all NodeName values associated to a given NodeHostname (useful
to get the list of virtual nodes associated with a real node in a configuration where multiple
slurmd daemons execute on a single compute node). assoc_mgr displays the current contents of the
slurmctld's internal cache for users, associations and/or qos. The ID may be
users=<user1>,[...,<userN>], accounts=<acct1>,[...,<acctN>], qos=<qos1>,[...,<qosN>] and/or
flags=<users,assoc,qos>, used to filter the desired section to be displayed. If no flags are
specified, all sections are displayed. burstbuffer displays the current status of the BurstBuffer
plugin. config displays parameter names from the configuration files in mixed case (e.g.
SlurmdPort=7003) while derived parameters names are in upper case only (e.g. SLURM_VERSION).
hostnames takes an optional hostlist expression as input and writes a list of individual host
names to standard output (one per line). If no hostlist expression is supplied, the contents of
the SLURM_JOB_NODELIST environment variable is used. For example "tux[1-3]" is mapped to
"tux1","tux2" and "tux3" (one hostname per line). hostlist takes a list of host names and prints
the hostlist expression for them (the inverse of hostnames). hostlist can also take the absolute
pathname of a file (beginning with the character '/') containing a list of hostnames. Multiple
node names may be specified using simple node range expressions (e.g. "lx[10-20]"). All other ID
values must identify a single element. The job step ID is of the form "job_id.step_id", (e.g.
"1234.1"). slurmd reports the current status of the slurmd daemon executing on the same node from
which the scontrol command is executed (the local host). It can be useful to diagnose problems.
By default hostlist does not sort the node list or make it unique (e.g. tux2,tux1,tux2 =
tux[2,1-2]). If you wanted a sorted list use hostlistsorted (e.g. tux2,tux1,tux2 = tux[1-2,2]).
By default, all elements of the entity type specified are printed. For an ENTITY of job, if the
job does not specify socket-per-node, cores-per-socket or threads-per-core then it will display
'*' in ReqS:C:T=*:*:* field. For an ENTITY of federation, the federation name that the controller
is part of and the sibling clusters part of the federation will be listed.
shutdown OPTION
Instruct Slurm daemons to save current state and terminate. By default, the Slurm controller
(slurmctld) forwards the request all other daemons (slurmd daemon on each compute node). An
OPTION of slurmctld or controller results in only the slurmctld daemon being shutdown and the
slurmd daemons remaining active.
suspend job_list
Suspend a running job. The job_list argument is a comma separated list of job IDs. Use the
resume command to resume its execution. User processes must stop on receipt of SIGSTOP signal and
resume upon receipt of SIGCONT for this operation to be effective. Not all architectures and
configurations support job suspension. If a suspended job is requeued, it will be placed in a
held state. The time a job is suspended will not count against a job's time limit. Only an
operator, administrator, SlurmUser, or root can suspend jobs.
takeover
Instruct Slurm's backup controller (slurmctld) to take over system control. Slurm's backup
controller requests control from the primary and waits for its termination. After that, it
switches from backup mode to controller mode. If primary controller can not be contacted, it
directly switches to controller mode. This can be used to speed up the Slurm controller fail-over
mechanism when the primary node is down. This can be used to minimize disruption if the computer
executing the primary Slurm controller is scheduled down. (Note: Slurm's primary controller will
take the control back at startup.)
top job_list
Move the specified job IDs to the top of the queue of jobs belonging to the identical user ID,
partition name, account, and QOS. The job_list argument is a comma separated ordered list of job
IDs. Any job not matching all of those fields will not be effected. Only jobs submitted to a
single partition will be effected. This operation changes the order of jobs by adjusting job nice
values. The net effect on that user's throughput will be negligible to slightly negative. This
operation is disabled by default for non-privileged (non-operator, admin, SlurmUser, or root)
users. This operation may be enabled for non-privileged users by the system administrator by
including the option "enable_user_top" in the SchedulerParameters configuration parameter.
uhold job_list
Prevent a pending job from being started (sets it's priority to 0). The job_list argument is a
space separated list of job IDs or job names. Use the release command to permit the job to be
scheduled. This command is designed for a system administrator to hold a job so that the job
owner may release it rather than requiring the intervention of a system administrator (also see
the hold command).
update SPECIFICATION
Update job, step, node, partition, powercapping or reservation configuration per the supplied
specification. SPECIFICATION is in the same format as the Slurm configuration file and the output
of the show command described above. It may be desirable to execute the show command (described
above) on the specific entity you want to update, then use cut-and-paste tools to enter updated
configuration values to the update. Note that while most configuration values can be changed using
this command, not all can be changed using this mechanism. In particular, the hardware
configuration of a node or the physical addition or removal of nodes from the cluster may only be
accomplished through editing the Slurm configuration file and executing the reconfigure command
(described above).
version
Display the version number of scontrol being executed.
wait_job job_id
Wait until a job and all of its nodes are ready for use or the job has entered some termination
state. This option is particularly useful in the Slurm Prolog or in the batch script itself if
nodes are powered down and restarted automatically as needed.
NOTE: Don't use scontrol wait_job in PrologCtld or Prolog with PrologFlags=Alloc as this will
result in deadlock.
write batch_script job_id optional_filename
Write the batch script for a given job_id to a file or to stdout. The file will default to
slurm-<job_id>.sh if the optional filename argument is not given. The script will be written to
stdout if - is given instead of a filename. The batch script can only be retrieved by an admin or
operator, or by the owner of the job.
write config
Write the current configuration to a file with the naming convention of "slurm.conf.<datetime>" in
the same directory as the original slurm.conf file.
INTERACTIVE COMMANDS
NOTE: All commands listed below can be used in the interactive mode, but NOT on the initial
command line.
all Show all partitions, their jobs and jobs steps. This causes information to be displayed about
partitions that are configured as hidden and partitions that are unavailable to user's group.
details
Causes the show command to provide additional details where available. Job information will
include CPUs and NUMA memory allocated on each node. Note that on computers with hyperthreading
enabled and Slurm configured to allocate cores, each listed CPU represents one physical core.
Each hyperthread on that core can be allocated a separate task, so a job's CPU count and task
count may differ. See the --cpu-bind and --mem-bind option descriptions in srun man pages for
more information. The details option is currently only supported for the show job command.
exit Terminate scontrol interactive session.
hide Do not display partition, job or jobs step information for partitions that are configured as
hidden or partitions that are unavailable to the user's group. This is the default behavior.
oneliner
Print information one line per record.
quiet Print no warning or informational messages, only fatal error messages.
quit Terminate the execution of scontrol.
verbose
Print detailed event logging. This includes time-stamps on data structures, record counts, etc.
!! Repeat the last command executed.
SPECIFICATIONS FOR UPDATE COMMAND, JOBS
Note that update requests done by either root, SlurmUser or Administrators are not subject to certain
restrictions. For instance, if an Administrator changes the QOS on a pending job, certain limits such as
the TimeLimit will not be changed automatically as changes made by the Administrators are allowed to
violate these restrictions.
Account=<account>
Account name to be changed for this job's resource use. Value may be cleared with blank data
value, "Account=".
AdminComment=<spec>
Arbitrary descriptive string. Can only be set by a Slurm administrator.
ArrayTaskThrottle=<count>
Specify the maximum number of tasks in a job array that can execute at the same time. Set the
count to zero in order to eliminate any limit. The task throttle count for a job array is
reported as part of its ArrayTaskId field, preceded with a percent sign. For example
"ArrayTaskId=1-10%2" indicates the maximum number of running tasks is limited to 2.
BurstBuffer=<spec>
Burst buffer specification to be changed for this job's resource use. Value may be cleared with
blank data value, "BurstBuffer=". Format is burst buffer plugin specific.
Clusters=<spec>
Specifies the clusters that the federated job can run on.
ClusterFeatures=<spec>
Specifies features that a federated cluster must have to have a sibling job submitted to it. Slurm
will attempt to submit a sibling job to a cluster if it has at least one of the specified
features.
Comment=<spec>
Arbitrary descriptive string.
Contiguous=<yes|no>
Set the job's requirement for contiguous (consecutive) nodes to be allocated. Possible values are
"YES" and "NO". Only the Slurm administrator or root can change this parameter.
CoreSpec=<count>
Number of cores to reserve per node for system use. The job will be charged for these cores, but
be unable to use them. Will be reported as "*" if not constrained.
CPUsPerTask=<count>
Change the CPUsPerTask job's value.
Deadline=<time_spec>
It accepts times of the form HH:MM:SS to specify a deadline to a job at a specific time of day
(seconds are optional). You may also specify midnight, noon, fika (3 PM) or teatime (4 PM) and
you can have a time-of-day suffixed with AM or PM for a deadline in the morning or the evening.
You can specify a deadline for the job with a date of the form MMDDYY or MM/DD/YY or MM.DD.YY, or
a date and time as YYYY-MM-DD[THH:MM[:SS]]. You can also give times like now + count time-units,
where the time-units can be minutes, hours, days, or weeks and you can tell Slurm to put a
deadline for tomorrow with the keyword tomorrow. The specified deadline must be later than the
current time. Only pending jobs can have the deadline updated. Only the Slurm administrator or
root can change this parameter.
DelayBoot=<time_spec>
Change the time to decide whether to reboot nodes in order to satisfy job's feature specification
if the job has been eligible to run for less than this time period. See salloc/sbatch man pages
option --delay-boot.
Dependency=<dependency_list>
Defer job's initiation until specified job dependency specification is satisfied. Cancel
dependency with an empty dependency_list (e.g. "Dependency="). <dependency_list> is of the form
<type:job_id[:job_id][,type:job_id[:job_id]]>. Many jobs can share the same dependency and these
jobs may even belong to different users.
after:job_id[:jobid...]
This job can begin execution after the specified jobs have begun execution.
afterany:job_id[:jobid...]
This job can begin execution after the specified jobs have terminated.
afternotok:job_id[:jobid...]
This job can begin execution after the specified jobs have terminated in some failed state
(non-zero exit code, node failure, timed out, etc).
afterok:job_id[:jobid...]
This job can begin execution after the specified jobs have successfully executed (ran to
completion with an exit code of zero).
singleton
This job can begin execution after any previously launched jobs sharing the same job name
and user have terminated. In other words, only one job by that name and owned by that user
can be running or suspended at any point in time.
EligibleTime=<time_spec>
See StartTime.
EndTime
The time the job is expected to terminate based on the job's time limit. When the job ends
sooner, this field will be updated with the actual end time.
ExcNodeList=<nodes>
Set the job's list of excluded node. Multiple node names may be specified using simple node range
expressions (e.g. "lx[10-20]"). Value may be cleared with blank data value, "ExcNodeList=".
Features=<features>
Set the job's required node features. The list of features may include multiple feature names
separated by ampersand (AND) and/or vertical bar (OR) operators. For example:
Features="opteron&video" or Features="fast|faster". In the first example, only nodes having both
the feature "opteron" AND the feature "video" will be used. There is no mechanism to specify that
you want one node with feature "opteron" and another node with feature "video" in case no node has
both features. If only one of a set of possible options should be used for all allocated nodes,
then use the OR operator and enclose the options within square brackets. For example:
"Features=[rack1|rack2|rack3|rack4]" might be used to specify that all nodes must be allocated on
a single rack of the cluster, but any of those four racks can be used. A request can also specify
the number of nodes needed with some feature by appending an asterisk and count after the feature
name. For example "Features=graphics*4" indicates that at least four allocated nodes must have
the feature "graphics." Parenthesis are also supported for features to be ANDed together. For
example "Features=[(knl&a2a&flat)*4&haswell*2]" indicates the resource allocation should include 4
nodes with ALL of the features "knl", "a2a", and "flat" plus 2 nodes with the feature "haswell".
Constraints with node counts may only be combined with AND operators. Value may be cleared with
blank data value, for example "Features=".
Gres=<list>
Specifies a comma delimited list of generic consumable resources. The format of each entry on the
list is "name[:count[*cpu]]". The name is that of the consumable resource. The count is the
number of those resources with a default value of 1. The specified resources will be allocated to
the job on each node allocated unless "*cpu" is appended, in which case the resources will be
allocated on a per cpu basis. The available generic consumable resources is configurable by the
system administrator. A list of available generic consumable resources will be printed and the
command will exit if the option argument is "help". Examples of use include
"Gres=gpus:2*cpu,disk=40G" and "Gres=help".
JobId=<job_list>
Identify the job(s) to be updated. The job_list may be a comma separated list of job IDs. Either
JobId or JobName is required.
Licenses=<name>
Specification of licenses (or other resources available on all nodes of the cluster) as described
in salloc/sbatch/srun man pages.
MinCPUsNode=<count>
Set the job's minimum number of CPUs per node to the specified value.
MinMemoryCPU=<megabytes>
Set the job's minimum real memory required per allocated CPU to the specified value. Either
MinMemoryCPU or MinMemoryNode may be set, but not both.
MinMemoryNode=<megabytes>
Set the job's minimum real memory required per node to the specified value. Either MinMemoryCPU
or MinMemoryNode may be set, but not both.
MinTmpDiskNode=<megabytes>
Set the job's minimum temporary disk space required per node to the specified value. Only the
Slurm administrator or root can change this parameter.
imeMin=<timespec>
Change TimeMin value which specifies the minimum time limit minutes of the job.
JobName=<name>
Identify the name of jobs to be modified or set the job's name to the specified value. When used
to identify jobs to be modified, all jobs belonging to all users are modified unless the UserID
option is used to identify a specific user. Either JobId or JobName is required.
Name[=<name>]
See JobName.
Nice[=<adjustment>]
Update the job with an adjusted scheduling priority within Slurm. With no adjustment value the
scheduling priority is decreased by 100. A negative nice value increases the priority, otherwise
decreases it. The adjustment range is +/- 2147483645. Only privileged users can specify a negative
adjustment.
NodeList=<nodes>
Change the nodes allocated to a running job to shrink it's size. The specified list of nodes must
be a subset of the nodes currently allocated to the job. Multiple node names may be specified
using simple node range expressions (e.g. "lx[10-20]"). After a job's allocation is reduced,
subsequent srun commands must explicitly specify node and task counts which are valid for the new
allocation.
NumCPUs=<min_count>[-<max_count>]
Set the job's minimum and optionally maximum count of CPUs to be allocated.
NumNodes=<min_count>[-<max_count>]
Set the job's minimum and optionally maximum count of nodes to be allocated. If the job is
already running, use this to specify a node count less than currently allocated and resources
previously allocated to the job will be relinquished. After a job's allocation is reduced,
subsequent srun commands must explicitly specify node and task counts which are valid for the new
allocation. Also see the NodeList parameter above. This is the same than ReqNodes.
NumTasks=<count>
Set the job's count of required tasks to the specified value. This is the same than ReqProcs.
OverSubscribe=<yes|no>
Set the job's ability to share compute resources (i.e. individual CPUs) with other jobs. Possible
values are "YES" and "NO". This option can only be changed for pending jobs.
Partition=<name>
Set the job's partition to the specified value.
Priority=<number>
Set the job's priority to the specified value. Note that a job priority of zero prevents the job
from ever being scheduled. By setting a job's priority to zero it is held. Set the priority to a
non-zero value to permit it to run. Explicitly setting a job's priority clears any previously set
nice value and removes the priority/multifactor plugin's ability to manage a job's priority. In
order to restore the priority/multifactor plugin's ability to manage a job's priority, hold and
then release the job. Only the Slurm administrator or root can increase job's priority.
QOS=<name>
Set the job's QOS (Quality Of Service) to the specified value. Value may be cleared with blank
data value, "QOS=".
Reboot=<yes|no>
Set the job's flag that specifies whether to force the allocated nodes to reboot before starting
the job. This is only supported with some system configurations and therefore it could be silently
ignored.
ReqCores=<count>
Change the job's requested Cores count.
ReqNodeList=<nodes>
Set the job's list of required node. Multiple node names may be specified using simple node range
expressions (e.g. "lx[10-20]"). Value may be cleared with blank data value, "ReqNodeList=".
ReqNodes=<min_count>[-<max_count>]
See NumNodes.
ReqProcs=<count>
See NumTasks.
ReqSockets=<count>
Change the job's requested socket count.
ReqThreads=<count>
Change the job's requested threads count.
Requeue=<0|1>
Stipulates whether a job should be requeued after a node failure: 0 for no, 1 for yes.
ReservationName=<name>
Set the job's reservation to the specified value. Value may be cleared with blank data value,
"ReservationName=".
ResetAccrueTime
Reset the job's accrue time value to 0 meaning it will loose any time previously accrued for
priority. Helpful if you have a large queue of jobs already in the queue and want to start
limiting how many jobs can accrue time without waiting for the queue to flush out.
SiteFactor=<account>
Specify the job's admin priority factor in the range of +/-2147483645. Only privileged users can
modify the value.
StdOut=<filepath>
Set the batch job's stdout file path.
Shared=<yes|no>
See OverSubscribe option above.
StartTime=<time_spec>
Set the job's earliest initiation time. It accepts times of the form HH:MM:SS to run a job at a
specific time of day (seconds are optional). (If that time is already past, the next day is
assumed.) You may also specify midnight, noon, fika (3 PM) or teatime (4 PM) and you can have a
time-of-day suffixed with AM or PM for running in the morning or the evening. You can also say
what day the job will be run, by specifying a date of the form MMDDYY or MM/DD/YY or MM.DD.YY, or
a date and time as YYYY-MM-DD[THH:MM[:SS]]. You can also give times like now + count time-units,
where the time-units can be minutes, hours, days, or weeks and you can tell Slurm to run the job
today with the keyword today and to run the job tomorrow with the keyword tomorrow.
Notes on date/time specifications:
- although the 'seconds' field of the HH:MM:SS time specification is allowed by the code, note
that the poll time of the Slurm scheduler is not precise enough to guarantee dispatch of the job
on the exact second. The job will be eligible to start on the next poll following the specified
time. The exact poll interval depends on the Slurm scheduler (e.g., 60 seconds with the default
sched/builtin).
- if no time (HH:MM:SS) is specified, the default is (00:00:00).
- if a date is specified without a year (e.g., MM/DD) then the current year is assumed, unless
the combination of MM/DD and HH:MM:SS has already passed for that year, in which case the next
year is used.
Switches=<count>[@<max-time-to-wait>]
When a tree topology is used, this defines the maximum count of switches desired for the job
allocation. If Slurm finds an allocation containing more switches than the count specified, the
job remain pending until it either finds an allocation with desired switch count or the time limit
expires. By default there is no switch count limit and no time limit delay. Set the count to zero
in order to clean any previously set count (disabling the limit). The job's maximum time delay
may be limited by the system administrator using the SchedulerParameters configuration parameter
with the max_switch_wait parameter option. Also see wait-for-switch.
wait-for-switch=<seconds>
Change max time to wait for a switch <seconds> secs.
TasksPerNode=<count>
Change the job's requested TasksPerNode.
ThreadSpec=<count>
Number of threads to reserve per node for system use. The job will be charged for these threads,
but be unable to use them. Will be reported as "*" if not constrained.
TimeLimit=<time>
The job's time limit. Output format is [days-]hours:minutes:seconds or "UNLIMITED". Input format
(for update command) set is minutes, minutes:seconds, hours:minutes:seconds, days-hours,
days-hours:minutes or days-hours:minutes:seconds. Time resolution is one minute and second values
are rounded up to the next minute. If changing the time limit of a job, either specify a new time
limit value or precede the time and equal sign with a "+" or "-" to increment or decrement the
current time limit (e.g. "TimeLimit+=30"). In order to increment or decrement the current time
limit, the JobId specification must precede the TimeLimit specification. Note that incrementing
or decrementing the time limit for a job array is only allowed before the job array has been split
into more than one job record. Only the Slurm administrator or root can increase job's TimeLimit.
UserID=<UID or name>
Used with the JobName option to identify jobs to be modified. Either a user name or numeric ID
(UID), may be specified.
WCKey=<key>
Set the job's workload characterization key to the specified value.
WorkDir=<directory_name>
Set the job's working directory to the specified value. Note that this may only be set for jobs in
the PENDING state, and that jobs may fail to launch if they rely on relative paths to the
originally submitted WorkDir.
SPECIFICATIONS FOR SHOW COMMAND, JOBS
The "show" command, when used with the "job" or "job <jobid>" entity displays detailed information about
a job or jobs. Much of this information may be modified using the "update job" command as described
above. However, the following fields displayed by the show job command are read-only and cannot be
modified:
AllocNode:Sid
Local node and system id making the resource allocation.
BatchFlag
Jobs submitted using the sbatch command have BatchFlag set to 1. Jobs submitted using other
commands have BatchFlag set to 0.
ExitCode=<exit>:<sig>
Exit status reported for the job by the wait() function. The first number is the exit code,
typically as set by the exit() function. The second number of the signal that caused the process
to terminate if it was terminated by a signal.
GroupId
The group under which the job was submitted.
JobState
The current state of the job.
NodeListIndices
The NodeIndices expose the internal indices into the node table associated with the node(s)
allocated to the job.
NtasksPerN:B:S:C=
<tasks_per_node>:<tasks_per_baseboard>:<tasks_per_socket>:<tasks_per_core> Specifies the number of
tasks to be started per hardware component (node, baseboard, socket and core). Unconstrained
values may be shown as "0" or "*".
PreemptEligibleTime
Time the job becomes eligible for preemption. Modified by PreemptExemptTime, either from the
global option in slurm.conf or the job QOS. This is hidden if the job has not started or if
PreemptMode=OFF.
PreemptTime
Time at which job was signaled that it was selected for preemption. (Meaningful only for
PreemptMode=CANCEL and the partition or QOS with which the job is associated has a GraceTime value
designated.) This is hidden if the job has not started or if PreemptMode=OFF.
PreSusTime
Time the job ran prior to last suspend.
Reason The reason job is not running: e.g., waiting "Resources".
ReqB:S:C:T=
<baseboard_count>:<socket_per_baseboard_count>:<core_per_socket_count>:<thread_per_core_count>
Specifies the count of various hardware components requested by the job. Unconstrained values may
be shown as "0" or "*".
SecsPreSuspend=<seconds>
If the job is suspended, this is the run time accumulated by the job (in seconds) prior to being
suspended.
Socks/Node=<count>
Count of desired sockets per node
SubmitTime
The time and date stamp (in localtime) the job was submitted. The format of the output is
identical to that of the EndTime field.
NOTE: If a job is requeued, the submit time is reset. To obtain the original submit time it is
necessary to use the "sacct -j <job_id[.<step_id>]" command also designating the -D or --duplicate
option to display all duplicate entries for a job.
SuspendTime
Time the job was last suspended or resumed.
NOTE on information displayed for various job states:
When you submit a request for the "show job" function the scontrol process makes an RPC request
call to slurmctld with a REQUEST_JOB_INFO message type. If the state of the job is PENDING, then
it returns some detail information such as: min_nodes, min_procs, cpus_per_task, etc. If the state
is other than PENDING the code assumes that it is in a further state such as RUNNING, COMPLETE,
etc. In these cases the code explicitly returns zero for these values. These values are
meaningless once the job resources have been allocated and the job has started.
SPECIFICATIONS FOR UPDATE COMMAND, STEPS
StepId=<job_id>[.<step_id>]
Identify the step to be updated. If the job_id is given, but no step_id is specified then all
steps of the identified job will be modified. This specification is required.
CompFile=<completion file>
Update a step with information about a steps completion. Can be useful if step statistics aren't
directly available through a jobacct_gather plugin. The file is a space-delimited file with
format for Version 1 is as follows
1 34461 0 2 0 3 1361906011 1361906015 1 1 3368 13357 /bin/sleep
A B C D E F G H I J K L M
Field Descriptions:
A file version
B ALPS apid
C inblocks
D outblocks
E exit status
F number of allocated CPUs
G start time
H end time
I utime
J stime
K maxrss
L uid
M command name
TimeLimit=<time>
The job's time limit. Output format is [days-]hours:minutes:seconds or "UNLIMITED". Input format
(for update command) set is minutes, minutes:seconds, hours:minutes:seconds, days-hours,
days-hours:minutes or days-hours:minutes:seconds. Time resolution is one minute and second values
are rounded up to the next minute. If changing the time limit of a step, either specify a new
time limit value or precede the time with a "+" or "-" to increment or decrement the current time
limit (e.g. "TimeLimit=+30"). In order to increment or decrement the current time limit, the
StepId specification must precede the TimeLimit specification.
SPECIFICATIONS FOR UPDATE COMMAND, NODES
NodeName=<name>
Identify the node(s) to be updated. Multiple node names may be specified using simple node range
expressions (e.g. "lx[10-20]"). This specification is required.
ActiveFeatures=<features>
Identify the feature(s) currently active on the specified node. Any previously active feature
specification will be overwritten with the new value. Also see AvailableFeatures. Typically
ActiveFeatures will be identical to AvailableFeatures; however ActiveFeatures may be configured as
a subset of the AvailableFeatures. For example, a node may be booted in multiple configurations.
In that case, all possible configurations may be identified as AvailableFeatures, while
ActiveFeatures would identify the current node configuration.
AvailableFeatures=<features>
Identify the feature(s) available on the specified node. Any previously defined available feature
specification will be overwritten with the new value. AvailableFeatures assigned via scontrol
will only persist across the restart of the slurmctld daemon with the -R option and state files
preserved or slurmctld's receipt of a SIGHUP. Update slurm.conf with any changes meant to be
persistent across normal restarts of slurmctld or the execution of scontrol reconfig. Also see
ActiveFeatures.
CpuBind=<node>
Specify the task binding mode to be used by default for this node. Supported options include:
"none", "board", "socket", "ldom" (NUMA), "core", "thread" and "off" (remove previous binding
mode).
Gres=<gres>
Identify generic resources to be associated with the specified node. Any previously defined
generic resources will be overwritten with the new value. Specifications for multiple generic
resources should be comma separated. Each resource specification consists of a name followed by
an optional colon with a numeric value (default value is one) (e.g. "Gres=bandwidth:10000").
Modification of GRES count associated with specific files (e.g. GPUs) is not allowed other than to
set their count on a node to zero. In order to change the GRES count to another value, modify
your slurm.conf and gres.conf files and restart daemons. If GRES as associated with specific
sockets, that information will be reported For example if all 4 GPUs on a node are all associated
with socket zero, then "Gres=gpu:4(S:0)". If associated with sockets 0 and 1 then
"Gres=gpu:4(S:0-1)". The information of which specific GPUs are associated with specific GPUs is
not reported, but only available by parsing the gres.conf file. Generic resources assigned via
scontrol will only persist across the restart of the slurmctld daemon with the -R option and state
files preserved or slurmctld's receipt of a SIGHUP. Update slurm.conf with any changes meant to
be persistent across normal restarts of slurmctld or the execution of scontrol reconfig.
Reason=<reason>
Identify the reason the node is in a "DOWN", "DRAINED", "DRAINING", "FAILING" or "FAIL" state.
Use quotes to enclose a reason having more than one word.
State=<state>
Identify the state to be assigned to the node. Possible node states are "NoResp", "ALLOC",
"ALLOCATED", "COMPLETING", "DOWN", "DRAIN", "FAIL", "FAILING", "FUTURE" "IDLE", "MAINT", "MIXED",
"PERFCTRS/NPC", "RESERVED", "POWER_DOWN", "POWER_UP", "RESUME" or "UNDRAIN". Not all of those
states can be set using the scontrol command only the following can: "CANCEL_REBOOT", "DOWN",
"DRAIN", "FAIL", "FUTURE", "RESUME", "NoResp", "POWER_DOWN", "POWER_UP" and "UNDRAIN". If a node
is in a "MIXED" state it usually means the node is in multiple states. For instance if only part
of the node is "ALLOCATED" and the rest of the node is "IDLE" the state will be "MIXED". If you
want to remove a node from service, you typically want to set it's state to "DRAIN".
"CANCEL_REBOOT" cancels a pending reboot on the node (same as "scontrol cancel_reboot <node>").
"FAILING" is similar to "DRAIN" except that some applications will seek to relinquish those nodes
before the job completes. "PERFCTRS/NPC" indicates that Network Performance Counters associated
with this node are in use, rendering this node as not usable for any other jobs. "RESERVED"
indicates the node is in an advanced reservation and not generally available. "RESUME" is not an
actual node state, but will change a node state from "DRAINED", "DRAINING", "DOWN" or "REBOOT" to
either "IDLE" or "ALLOCATED" state as appropriate. "RESUME" will also clear the "POWERING_DOWN"
state of a node and make it eligible to be allocted. "UNDRAIN" clears the node from being drained
(like "RESUME"), but will not change the node's base state (e.g. "DOWN"). Setting a node "DOWN"
will cause all running and suspended jobs on that node to be terminated. "POWER_DOWN" and
"POWER_UP" will use the configured SuspendProg and ResumeProg programs to explicitly place a node
in or out of a power saving mode. If a node is already in the process of being powered up or down,
the command will only change the state of the node but won't have any effect until the configured
ResumeTimeout or SuspendTimeout is reached. Use of this command can be useful in situations where
a ResumeProg like capmc in Cray machines is stalled and one wants to restore the node to "IDLE"
manually, in this case rebooting the node and setting the state to "POWER_DOWN" will cancel the
previous "POWER_UP" state and the node will become "IDLE". The "NoResp" state will only set the
"NoResp" flag for a node without changing its underlying state. While all of the above states are
valid, some of them are not valid new node states given their prior state. If the node state code
printed is followed by "~", this indicates the node is presently in a power saving mode (typically
running at reduced frequency). If the node state code is followed by "#", this indicates the node
is presently being powered up or configured. If the node state code is followed by "$", this
indicates the node is currently in a reservation with a flag value of "maintenance". If the node
state code is followed by "@", this indicates the node is currently scheduled to be rebooted.
Generally only "DRAIN", "FAIL" and "RESUME" should be used. NOTE: The scontrol command should not
be used to change node state on Cray systems. Use Cray tools such as xtprocadmin instead.
Weight=<weight>
Identify weight to be associated with specified nodes. This allows dynamic changes to weight
associated with nodes, which will be used for the subsequent node allocation decisions. Weight
assigned via scontrol will only persist across the restart of the slurmctld daemon with the -R
option and state files preserved or slurmctld's receipt of a SIGHUP. Update slurm.conf with any
changes meant to be persistent across normal restarts of slurmctld or the execution of scontrol
reconfig.
SPECIFICATIONS FOR SHOW COMMAND, NODES
The meaning of the energy information is as follows:
CurrentWatts
The instantaneous power consumption of the node at the time of the last node energy accounting
sample, in watts.
LowestJoules
The energy consumed by the node between the last time it was powered on and the last time it was
registered by slurmd, in joules.
ConsumedJoules
The energy consumed by the node between the last time it was registered by the slurmd daemon and
the last node energy accounting sample, in joules.
If the reported value is "n/s" (not supported), the node does not support the configured
AcctGatherEnergyType plugin. If the reported value is zero, energy accounting for nodes is disabled.
The meaning of the external sensors information is as follows:
ExtSensorsJoules
The energy consumed by the node between the last time it was powered on and the last external
sensors plugin node sample, in joules.
ExtSensorsWatts
The instantaneous power consumption of the node at the time of the last external sensors plugin
node sample, in watts.
ExtSensorsTemp
The temperature of the node at the time of the last external sensors plugin node sample, in
celsius.
If the reported value is "n/s" (not supported), the node does not support the configured ExtSensorsType
plugin.
The meaning of the resource specialization information is as follows:
CPUSpecList
The list of Slurm abstract CPU IDs on this node reserved for exclusive use by the Slurm compute
node daemons (slurmd, slurmstepd).
MemSpecLimit
The combined memory limit, in megabytes, on this node for the Slurm compute node daemons (slurmd,
slurmstepd).
The meaning of the memory information is as follows:
RealMemory
The total memory, in MB, on the node.
AllocMem
The total memory, in MB, currently allocated by jobs on the node.
FreeMem
The total memory, in MB, currently free on the node as reported by the OS.
SPECIFICATIONS FOR UPDATE COMMAND, FRONTEND
FrontendName=<name>
Identify the front end node to be updated. This specification is required.
Reason=<reason>
Identify the reason the node is in a "DOWN" or "DRAIN" state. Use quotes to enclose a reason
having more than one word.
State=<state>
Identify the state to be assigned to the front end node. Possible values are "DOWN", "DRAIN" or
"RESUME". If you want to remove a front end node from service, you typically want to set it's
state to "DRAIN". "RESUME" is not an actual node state, but will return a "DRAINED", "DRAINING",
or "DOWN" front end node to service, either "IDLE" or "ALLOCATED" state as appropriate. Setting a
front end node "DOWN" will cause all running and suspended jobs on that node to be terminated.
SPECIFICATIONS FOR CREATE, UPDATE, AND DELETE COMMANDS, PARTITIONS
AllowGroups=<name>
Identify the user groups which may use this partition. Multiple groups may be specified in a
comma separated list. To permit all groups to use the partition specify "AllowGroups=ALL".
AllocNodes=<name>
Comma separated list of nodes from which users can execute jobs in the partition. Node names may
be specified using the node range expression syntax described above. The default value is "ALL".
Alternate=<partition name>
Alternate partition to be used if the state of this partition is "DRAIN" or "INACTIVE." The value
"NONE" will clear a previously set alternate partition.
CpuBind=<node>
Specify the task binding mode to be used by default for this partition. Supported options
include: "none", "board", "socket", "ldom" (NUMA), "core", "thread" and "off" (remove previous
binding mode).
Default=<yes|no>
Specify if this partition is to be used by jobs which do not explicitly identify a partition to
use. Possible output values are "YES" and "NO". In order to change the default partition of a
running system, use the scontrol update command and set Default=yes for the partition that you
want to become the new default.
DefaultTime=<time>
Run time limit used for jobs that don't specify a value. If not set then MaxTime will be used.
Format is the same as for MaxTime.
DefMemPerCPU=<MB>
Set the default memory to be allocated per CPU for jobs in this partition. The memory size is
specified in megabytes.
DefMemPerNode=<MB>
Set the default memory to be allocated per node for jobs in this partition. The memory size is
specified in megabytes.
DisableRootJobs=<yes|no>
Specify if jobs can be executed as user root. Possible values are "YES" and "NO".
GraceTime=<seconds>
Specifies, in units of seconds, the preemption grace time to be extended to a job which has been
selected for preemption. The default value is zero, no preemption grace time is allowed on this
partition or qos. (Meaningful only for PreemptMode=CANCEL)
Hidden=<yes|no>
Specify if the partition and its jobs should be hidden from view. Hidden partitions will by
default not be reported by Slurm APIs or commands. Possible values are "YES" and "NO".
JobDefaults=<specs>
Specify job default values using a comma delimited list of "key=value" pairs. Supported keys
include
DefCpuPerGPU Default number of CPUs per allocated GPU.
DefMemPerGPU Default memory limit (in megabytes) per allocated GPU.
MaxMemPerCPU=<MB>
Set the maximum memory to be allocated per CPU for jobs in this partition. The memory size is
specified in megabytes.
MaxMemPerCNode=<MB>
Set the maximum memory to be allocated per node for jobs in this partition. The memory size is
specified in megabytes.
MaxNodes=<count>
Set the maximum number of nodes which will be allocated to any single job in the partition.
Specify a number, "INFINITE" or "UNLIMITED". Changing the MaxNodes of a partition has no effect
upon jobs that have already begun execution.
MaxTime=<time>
The maximum run time for jobs. Output format is [days-]hours:minutes:seconds or "UNLIMITED".
Input format (for update command) is minutes, minutes:seconds, hours:minutes:seconds, days-hours,
days-hours:minutes or days-hours:minutes:seconds. Time resolution is one minute and second values
are rounded up to the next minute. Changing the MaxTime of a partition has no effect upon jobs
that have already begun execution.
MinNodes=<count>
Set the minimum number of nodes which will be allocated to any single job in the partition.
Changing the MinNodes of a partition has no effect upon jobs that have already begun execution.
Nodes=<name>
Identify the node(s) to be associated with this partition. Multiple node names may be specified
using simple node range expressions (e.g. "lx[10-20]"). Note that jobs may only be associated
with one partition at any time. Specify a blank data value to remove all nodes from a partition:
"Nodes=". Changing the Nodes in a partition has no effect upon jobs that have already begun
execution.
OverTimeLimit=<count>
Number of minutes by which a job can exceed its time limit before being canceled. The configured
job time limit is treated as a soft limit. Adding OverTimeLimit to the soft limit provides a hard
limit, at which point the job is canceled. This is particularly useful for backfill scheduling,
which bases upon each job's soft time limit. A partition-specific OverTimeLimit will override any
global OverTimeLimit value. If not specified, the global OverTimeLimit value will take
precedence. May not exceed exceed 65533 minutes. An input value of "UNLIMITED" will clear any
previously configured partition-specific OverTimeLimit value.
OverSubscribe=<yes|no|exclusive|force>[:<job_count>]
Specify if compute resources (i.e. individual CPUs) in this partition can be shared by multiple
jobs. Possible values are "YES", "NO", "EXCLUSIVE" and "FORCE". An optional job count specifies
how many jobs can be allocated to use each resource.
PartitionName=<name>
Identify the partition to be updated. This specification is required.
PreemptMode=<mode>
Reset the mechanism used to preempt jobs in this partition if PreemptType is configured to
preempt/partition_prio. The default preemption mechanism is specified by the cluster-wide
PreemptMode configuration parameter. Possible values are "OFF", "CANCEL", "CHECKPOINT", "REQUEUE"
and "SUSPEND".
Priority=<count>
Jobs submitted to a higher priority partition will be dispatched before pending jobs in lower
priority partitions and if possible they will preempt running jobs from lower priority partitions.
Note that a partition's priority takes precedence over a job's priority. The value may not exceed
65533.
PriorityJobFactor=<count>
Partition factor used by priority/multifactor plugin in calculating job priority. The value may
not exceed 65533. Also see PriorityTier.
PriorityTier=<count>
Jobs submitted to a partition with a higher priority tier value will be dispatched before pending
jobs in partition with lower priority tier value and, if possible, they will preempt running
jobs from partitions with lower priority tier values. Note that a partition's priority tier takes
precedence over a job's priority. The value may not exceed 65533. Also see PriorityJobFactor.
QOS=<QOSname|blank to remove>
Set the partition QOS with a QOS name or to remove the Partition QOS leave the option blank.
RootOnly=<yes|no>
Specify if only allocation requests initiated by user root will be satisfied. This can be used to
restrict control of the partition to some meta-scheduler. Possible values are "YES" and "NO".
ReqResv=<yes|no>
Specify if only allocation requests designating a reservation will be satisfied. This is used to
restrict partition usage to be allowed only within a reservation. Possible values are "YES" and
"NO".
Shared=<yes|no|exclusive|force>[:<job_count>]
Renamed to OverSubscribe, see option descriptions above.
State=<up|down|drain|inactive>
Specify if jobs can be allocated nodes or queued in this partition. Possible values are "UP",
"DOWN", "DRAIN" and "INACTIVE".
UP Designates that new jobs may queued on the partition, and that jobs may be allocated
nodes and run from the partition.
DOWN Designates that new jobs may be queued on the partition, but queued jobs may not be
allocated nodes and run from the partition. Jobs already running on the partition
continue to run. The jobs must be explicitly canceled to force their termination.
DRAIN Designates that no new jobs may be queued on the partition (job submission requests will
be denied with an error message), but jobs already queued on the partition may be
allocated nodes and run. See also the "Alternate" partition specification.
INACTIVE Designates that no new jobs may be queued on the partition, and jobs already queued may
not be allocated nodes and run. See also the "Alternate" partition specification.
TRESBillingWeights=<TRES Billing Weights>
TRESBillingWeights is used to define the billing weights of each TRES type that will be used in
calculating the usage of a job. The calculated usage is used when calculating fairshare and when
enforcing the TRES billing limit on jobs. Updates affect new jobs and not existing jobs. See the
slurm.conf man page for more information.
SPECIFICATIONS FOR UPDATE COMMAND, POWERCAP
PowerCap=<count>
Set the amount of watts the cluster is limited to. Specify a number, "INFINITE" to enable the
power capping logic without power restriction or "0" to disable the power capping logic. Update
slurm.conf with any changes meant to be persistent across normal restarts of slurmctld or the
execution of scontrol reconfig.
SPECIFICATIONS FOR CREATE, UPDATE, AND DELETE COMMANDS, RESERVATIONS
Reservation=<name>
Identify the name of the reservation to be created, updated, or deleted. This parameter is
required for update and is the only parameter for delete. For create, if you do not want to give
a reservation name, use "scontrol create res ..." and a name will be created automatically.
Accounts=<account list>
List of accounts permitted to use the reserved nodes, for example "Accounts=physcode1,physcode2".
A user in any of the accounts may use the reserved nodes. A new reservation must specify Users
and/or Accounts. If both Users and Accounts are specified, a job must match both in order to use
the reservation. Accounts can also be denied access to reservations by preceding all of the
account names with '-'. Alternately precede the equal sign with '-'. For example,
"Accounts=-physcode1,-physcode2" or "Accounts-=physcode1,physcode2" will permit any account except
physcode1 and physcode2 to use the reservation. You can add or remove individual accounts from an
existing reservation by using the update command and adding a '+' or '-' sign before the '=' sign.
If accounts are denied access to a reservation (account name preceded by a '-'), then all other
accounts are implicitly allowed to use the reservation and it is not possible to also explicitly
specify allowed accounts.
BurstBuffer=<buffer_spec>[,<buffer_spec>,...]
Specification of burst buffer resources which are to be reserved. "buffer_spec" consists of four
elements: [plugin:][type:]#[units] "plugin" is the burst buffer plugin name, currently either
"datawarp" or "generic". If no plugin is specified, the reservation applies to all configured
burst buffer plugins. "type" specifies a Cray generic burst buffer resource, for example "nodes".
if "type" is not specified, the number is a measure of storage space. The "units" may be "N"
(nodes), "K|KiB", "M|MiB", "G|GiB", "T|TiB", "P|PiB" (for powers of 1024) and "KB", "MB", "GB",
"TB", "PB" (for powers of 1000). The default units are bytes for reservations of storage space.
For example "BurstBuffer=datawarp:2TB" (reserve 2TB of storage plus 3 nodes from the Cray plugin)
or "BurstBuffer=100GB" (reserve 100 GB of storage from all configured burst buffer plugins). Jobs
using this reservation are not restricted to these burst buffer resources, but may use these
reserved resources plus any which are generally available. NOTE: Usually Slurm interprets KB, MB,
GB, TB, PB, TB units as powers of 1024, but for Burst Buffers size specifications Slurm supports
both IEC/SI formats. This is because the CRAY API for managing DataWarps supports both formats.
CoreCnt=<num>
This option is only supported when SelectType=select/cons_res. Identify number of cores to be
reserved. If NodeCnt is used without the FIRST_CORES flag, this is the total number of cores to
reserve where cores per node is CoreCnt/NodeCnt. If a nodelist is used, or if NodeCnt is used
with the FIRST_CORES flag, this should be an array of core numbers by node: Nodes=node[1-5]
CoreCnt=2,2,3,3,4 or flags=FIRST_CORES NodeCnt=5 CoreCnt=1,2,1,3,2.
Licenses=<license>
Specification of licenses (or other resources available on all nodes of the cluster) which are to
be reserved. License names can be followed by a colon and count (the default count is one).
Multiple license names should be comma separated (e.g. "Licenses=foo:4,bar"). A new reservation
must specify one or more resource to be included: NodeCnt, Nodes and/or Licenses. If a
reservation includes Licenses, but no NodeCnt or Nodes, then the option Flags=LICENSE_ONLY must
also be specified. Jobs using this reservation are not restricted to these licenses, but may use
these reserved licenses plus any which are generally available.
NodeCnt=<num>[,num,...]
Identify number of nodes to be reserved. The number can include a suffix of "k" or "K", in which
case the number specified is multiplied by 1024. A new reservation must specify one or more
resource to be included: NodeCnt, Nodes and/or Licenses.
Nodes=<name>
Identify the node(s) to be reserved. Multiple node names may be specified using simple node range
expressions (e.g. "Nodes=lx[10-20]"). Specify a blank data value to remove all nodes from a
reservation: "Nodes=". A new reservation must specify one or more resource to be included:
NodeCnt, Nodes and/or Licenses. A specification of "ALL" will reserve all nodes. Set
Flags=PART_NODES and PartitionName= in order for changes in the nodes associated with a partition
to also be reflected in the nodes associated with a reservation.
StartTime=<time_spec>
The start time for the reservation. A new reservation must specify a start time. It accepts
times of the form HH:MM:SS for a specific time of day (seconds are optional). (If that time is
already past, the next day is assumed.) You may also specify midnight, noon, fika (3 PM) or
teatime (4 PM) and you can have a time-of-day suffixed with AM or PM for running in the morning or
the evening. You can also say what day the job will be run, by specifying a date of the form
MMDDYY or MM/DD/YY or MM.DD.YY, or a date and time as YYYY-MM-DD[THH:MM[:SS]]. You can also give
times like now + count time-units, where the time-units can be minutes, hours, days, or weeks and
you can tell Slurm to run the job today with the keyword today and to run the job tomorrow with
the keyword tomorrow. You cannot update the StartTime of a reservation in ACTIVE state.
EndTime=<time_spec>
The end time for the reservation. A new reservation must specify an end time or a duration.
Valid formats are the same as for StartTime.
Duration=<time>
The length of a reservation. A new reservation must specify an end time or a duration. Valid
formats are minutes, minutes:seconds, hours:minutes:seconds, days-hours, days-hours:minutes,
days-hours:minutes:seconds, or UNLIMITED. Time resolution is one minute and second values are
rounded up to the next minute. Output format is always [days-]hours:minutes:seconds.
PartitionName=<name>
Identify the partition to be reserved.
Flags=<flags>
Flags associated with the reservation. You can add or remove individual flags from an existing
reservation by adding a '+' or '-' sign before the '=' sign. For example: Flags-=DAILY (NOTE:
this shortcut is not supported for all flags). Currently supported flags include:
ANY_NODES This is a reservation for burst buffers and/or licenses only and not compute nodes.
If this flag is set, a job using this reservation may use the associated burst
buffers and/or licenses plus any compute nodes. If this flag is not set, a job
using this reservation may use only the nodes and licenses associated with the
reservation.
DAILY Repeat the reservation at the same time every day.
FLEX Permit jobs requesting the reservation to begin prior to the reservation's start
time, end after the reservation's end time, and use any resources inside and/or
outside of the reservation regardless of any constraints possibly set in the
reservation. A typical use case is to prevent jobs not explicitly requesting the
reservation from using those reserved resources rather than forcing jobs requesting
the reservation to use those resources in the time frame reserved. Another use case
could be to always have a particular number of nodes with a specific feature
reserved for a specific account so users in this account may use this nodes plus
possibly other nodes without this feature.
FIRST_CORES Use the lowest numbered cores on a node only.
IGNORE_JOBS Ignore currently running jobs when creating the reservation. This can be especially
useful when reserving all nodes in the system for maintenance.
LICENSE_ONLY See ANY_NODES.
MAINT Maintenance mode, receives special accounting treatment. This partition is
permitted to use resources that are already in another reservation.
NO_HOLD_JOBS_AFTER
By default, when a reservation ends the reservation request will be removed from any
pending jobs submitted to the reservation and will be put into a held state. Use
this flag to let jobs run outside of the reservation after the reservation is gone.
OVERLAP This reservation can be allocated resources that are already in another reservation.
PART_NODES This flag can be used to reserve all nodes within the specified partition.
PartitionName and Nodes=ALL must be specified or this option is ignored.
PURGE_COMP Purge the reservation once the last associated job has completed. Once the
reservation has been created, it must be populated within 5 minutes of its start
time or it will be purged before any jobs have been run.
REPLACE Nodes which are DOWN, DRAINED, or allocated to jobs are automatically replenished
using idle resources. This option can be used to maintain a constant number of idle
resources available for pending jobs (subject to availability of idle resources).
This should be used with the NodeCnt reservation option; do not identify specific
nodes to be included in the reservation.
REPLACE_DOWN Nodes which are DOWN or DRAINED are automatically replenished using idle resources.
This option can be used to maintain a constant sized pool of resources available for
pending jobs (subject to availability of idle resources). This should be used with
the NodeCnt reservation option; do not identify specific nodes to be included in the
reservation.
SPEC_NODES Reservation is for specific nodes (output only)
STATIC_ALLOC Make it so after the nodes are selected for a reservation they don't change.
Without this option when nodes are selected for a reservation and one goes down the
reservation will select a new node to fill the spot.
TIME_FLOAT The reservation start time is relative to the current time and moves forward through
time (e.g. a StartTime=now+10minutes will always be 10 minutes in the future).
WEEKDAY Repeat the reservation at the same time on every weekday (Monday, Tuesday,
Wednesday, Thursday and Friday).
WEEKEND Repeat the reservation at the same time on every weekend day (Saturday and Sunday).
WEEKLY Repeat the reservation at the same time every week.
Features=<features>
Set the reservation's required node features. Multiple values may be "&" separated if all features
are required (AND operation) or separated by "|" if any of the specified features are required (OR
operation). Parenthesis are also supported for features to be ANDed together with counts of nodes
having the specified features. For example "Features=[(knl&a2a&flat)*4&haswell*2]" indicates the
advanced reservation should include 4 nodes with ALL of the features "knl", "a2a", and "flat" plus
2 nodes with the feature "haswell".
Value may be cleared with blank data value, "Features=".
Users=<user list>
List of users permitted to use the reserved nodes, for example "User=jones1,smith2". A new
reservation must specify Users and/or Accounts. If both Users and Accounts are specified, a job
must match both in order to use the reservation. Users can also be denied access to reservations
by preceding all of the user names with '-'. Alternately precede the equal sign with '-'. For
example, "User=-jones1,-smith2" or "User-=jones1,smith2" will permit any user except jones1 and
smith2 to use the reservation. You can add or remove individual users from an existing
reservation by using the update command and adding a '+' or '-' sign before the '=' sign. If
users are denied access to a reservation (user name preceded by a '-'), then all other users are
implicitly allowed to use the reservation and it is not possible to also explicitly specify
allowed users.
TRES=<tres_spec>
Comma-separated list of TRES required for the reservation. Current supported TRES types with
reservations are: CPU, Node, License and BB. CPU and Node follow the same format as CoreCnt and
NodeCnt parameters respectively. License names can be followed by an equal '=' and a count:
License/<name1>=<count1>[,License/<name2>=<count2>,...]
BurstBuffer can be specified in a similar way as BurstBuffer parameter. The only difference is
that colon symbol ':' should be replaced by an equal '=' in order to follow the TRES format.
Some examples of TRES valid specifications:
TRES=cpu=5,bb/cray=4,license/iop1=1,license/iop2=3
TRES=node=5k,license/iop1=2
As specified in CoreCnt, if a nodelist is specified, cpu can be an array of core numbers by node:
nodes=compute[1-3] TRES=cpu=2,2,1,bb/cray=4,license/iop1=2
Please note that CPU, Node, License and BB can override CoreCnt, NodeCnt, Licenses and BurstBuffer
parameters respectively. Also CPU represents CoreCnt, in a reservation and will be adjusted if
you have threads per core on your nodes.
SPECIFICATIONS FOR SHOW COMMAND, LAYOUTS
Without options, lists all configured layouts. With a layout specified, shows entities with following
options:
Key=<value>
Keys/Values to update for the entities. The format must respect the layout.d configuration files.
Key=Type cannot be updated. One Key/Value is required, several can be set.
Entity=<value>
Entities to show, default is not used. Can be set to "*".
Type=<value>
Type of entities to show, default is not used.
nolayout
If not used, only entities with defining the tree are shown. With the option, only leaves are
shown.
ENVIRONMENT VARIABLES
Some scontrol options may be set via environment variables. These environment variables, along with their
corresponding options, are listed below. (Note: Commandline options will always override these settings.)
SCONTROL_ALL -a, --all
SCONTROL_FEDERATION --federation
SCONTROL_FUTURE -F, --future
SCONTROL_LOCAL --local
SCONTROL_SIBLING --sibling
SLURM_BITSTR_LEN Specifies the string length to be used for holding a job array's task ID expression.
The default value is 64 bytes. A value of 0 will print the full expression with any
length required. Larger values may adversely impact the application performance.
SLURM_CLUSTERS Same as --clusters
SLURM_CONF The location of the Slurm configuration file.
SLURM_TIME_FORMAT Specify the format used to report time stamps. A value of standard, the default
value, generates output in the form "year-month-dateThour:minute:second". A value of
relative returns only "hour:minute:second" if the current day. For other dates in
the current year it prints the "hour:minute" preceded by "Tomorr" (tomorrow),
"Ystday" (yesterday), the name of the day for the coming week (e.g. "Mon", "Tue",
etc.), otherwise the date (e.g. "25 Apr"). For other years it returns a date month
and year without a time (e.g. "6 Jun 2012"). All of the time stamps use a 24 hour
format.
A valid strftime() format can also be specified. For example, a value of "%a %T" will
report the day of the week and a time stamp (e.g. "Mon 12:34:56").
SLURM_TOPO_LEN Specify the maximum size of the line when printing Topology. If not set, the default
value is unlimited.
AUTHORIZATION
When using the Slurm db, users who have AdminLevel's defined (Operator or Admin) and users who are
account coordinators are given the authority to view and modify jobs, reservations, nodes, etc., as
defined in the following table - regardless of whether a PrivateData restriction has been defined in the
slurm.conf file.
scontrol show job(s): Admin, Operator, Coordinator
scontrol update job: Admin, Operator, Coordinator
scontrol requeue: Admin, Operator, Coordinator
scontrol show step(s): Admin, Operator, Coordinator
scontrol update step: Admin, Operator, Coordinator
scontrol show node: Admin, Operator
scontrol update node: Admin
scontrol create partition: Admin
scontrol show partition: Admin, Operator
scontrol update partition: Admin
scontrol delete partition: Admin
scontrol create reservation: Admin, Operator
scontrol show reservation: Admin, Operator
scontrol update reservation: Admin, Operator
scontrol delete reservation: Admin, Operator
scontrol reconfig: Admin
scontrol shutdown: Admin
scontrol takeover: Admin
EXAMPLES
# scontrol
scontrol: show part debug
PartitionName=debug
AllocNodes=ALL AllowGroups=ALL Default=YES
DefaultTime=NONE DisableRootJobs=NO Hidden=NO
MaxNodes=UNLIMITED MaxTime=UNLIMITED MinNodes=1
Nodes=snowflake[0-48]
Priority=1 RootOnly=NO OverSubscribe=YES:4
State=UP TotalCPUs=694 TotalNodes=49
scontrol: update PartitionName=debug MaxTime=60:00 MaxNodes=4
scontrol: show job 71701
JobId=71701 Name=hostname
UserId=da(1000) GroupId=da(1000)
Priority=66264 Account=none QOS=normal WCKey=*123
JobState=COMPLETED Reason=None Dependency=(null)
TimeLimit=UNLIMITED Requeue=1 Restarts=0 BatchFlag=0 ExitCode=0:0
SubmitTime=2010-01-05T10:58:40 EligibleTime=2010-01-05T10:58:40
StartTime=2010-01-05T10:58:40 EndTime=2010-01-05T10:58:40
SuspendTime=None SecsPreSuspend=0
Partition=debug AllocNode:Sid=snowflake:4702
ReqNodeList=(null) ExcNodeList=(null)
NodeList=snowflake0
NumNodes=1 NumCPUs=10 CPUs/Task=2 ReqS:C:T=1:1:1
MinCPUsNode=2 MinMemoryNode=0 MinTmpDiskNode=0
Features=(null) Reservation=(null)
OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null)
scontrol: update JobId=71701 TimeLimit=30:00 Priority=500
scontrol: show hostnames tux[1-3]
tux1
tux2
tux3
scontrol: create res StartTime=2009-04-01T08:00:00 Duration=5:00:00 Users=dbremer NodeCnt=10
Reservation created: dbremer_1
scontrol: update Reservation=dbremer_1 Flags=Maint NodeCnt=20
scontrol: delete Reservation=dbremer_1
scontrol: quit
COPYING
Copyright (C) 2002-2007 The Regents of the University of California. Produced at Lawrence Livermore
National Laboratory (cf, DISCLAIMER).
Copyright (C) 2008-2010 Lawrence Livermore National Security.
Copyright (C) 2010-2018 SchedMD LLC.
This file is part of Slurm, a resource management program. For details, see
<https://slurm.schedmd.com/>.
Slurm is free software; you can redistribute it and/or modify it under the terms of the GNU General
Public License as published by the Free Software Foundation; either version 2 of the License, or (at your
option) any later version.
Slurm is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the
implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
License for more details.
FILES
/etc/slurm.conf
SEE ALSO
scancel(1), sinfo(1), squeue(1), slurm_checkpoint (3), slurm_create_partition (3), slurm_delete_partition
(3), slurm_load_ctl_conf (3), slurm_load_jobs (3), slurm_load_node (3), slurm_load_partitions (3),
slurm_reconfigure (3), slurm_requeue (3), slurm_resume (3), slurm_shutdown (3), slurm_suspend (3),
slurm_takeover (3), slurm_update_job (3), slurm_update_node (3), slurm_update_partition (3),
slurm.conf(5), slurmctld(8)