1scontrol(1) Slurm Commands scontrol(1)
2
3
4
6 scontrol - Used view and modify Slurm configuration and state.
7
8
10 scontrol [OPTIONS...] [COMMAND...]
11
12
14 scontrol is used to view or modify Slurm configuration including: job,
15 job step, node, partition, reservation, and overall system configura‐
16 tion. Most of the commands can only be executed by user root or an
17 Administrator. If an attempt to view or modify configuration informa‐
18 tion is made by an unauthorized user, an error message will be printed
19 and the requested action will not occur. If no command is entered on
20 the execute line, scontrol will operate in an interactive mode and
21 prompt for input. It will continue prompting for input and executing
22 commands until explicitly terminated. If a command is entered on the
23 execute line, scontrol will execute that command and terminate. All
24 commands and options are case-insensitive, although node names, parti‐
25 tion names, and reservation names are case-sensitive (node names "LX"
26 and "lx" are distinct). All commands and options can be abbreviated to
27 the extent that the specification is unique. A modified Slurm configu‐
28 ration can be written to a file using the scontrol write config com‐
29 mand. The resulting file will be named using the convention
30 "slurm.conf.<datetime>" and located in the same directory as the origi‐
31 nal "slurm.conf" file. The directory containing the original slurm.conf
32 must be writable for this to occur.
33
34
36 -a, --all
37 When the show command is used, then display all partitions,
38 their jobs and jobs steps. This causes information to be dis‐
39 played about partitions that are configured as hidden and parti‐
40 tions that are unavailable to user's group.
41
42 -d, --details
43 Causes the show command to provide additional details where
44 available.
45
46 --federation
47 Report jobs from from federation if a member of one.
48
49 -F, --future
50 Report nodes in FUTURE state.
51
52 -h, --help
53 Print a help message describing the usage of scontrol.
54
55 --hide Do not display information about hidden partitions, their jobs
56 and job steps. By default, neither partitions that are config‐
57 ured as hidden nor those partitions unavailable to user's group
58 will be displayed (i.e. this is the default behavior).
59
60 --local
61 Show only information local to this cluster. Ignore other clus‐
62 ters in the federated if a member of one. Overrides --federa‐
63 tion.
64
65 -M, --clusters=<string>
66 The cluster to issue commands to. Only one cluster name may be
67 specified. Note that the SlurmDBD must be up for this option to
68 work properly. This option implicitly sets the --local option.
69
70
71 -o, --oneliner
72 Print information one line per record.
73
74 -Q, --quiet
75 Print no warning or informational messages, only fatal error
76 messages.
77
78 --sibling
79 Show all sibling jobs on a federated cluster. Implies --federa‐
80 tion.
81
82 -u, --uid=<uid>
83 Attempt to update a job as user <uid> instead of the invoking
84 user id.
85
86 -v, --verbose
87 Print detailed event logging. Multiple -v's will further
88 increase the verbosity of logging. By default only errors will
89 be displayed.
90
91
92 -V , --version
93 Print version information and exit.
94
95 COMMANDS
96
97
98 abort Instruct the Slurm controller to terminate immediately and gen‐
99 erate a core file. See "man slurmctld" for information about
100 where the core file will be written.
101
102
103 cancel_reboot <NodeList>
104 Cancel pending reboots on nodes.
105
106
107 checkpoint CKPT_OP ID
108 Perform a checkpoint activity on the job step(s) with the speci‐
109 fied identification. ID can be used to identify a specific job
110 (e.g. "<job_id>", which applies to all of its existing steps) or
111 a specific job step (e.g. "<job_id>.<step_id>"). Acceptable
112 values for CKPT_OP include:
113
114 able Test if presently not disabled, report start time if
115 checkpoint in progress
116
117 create Create a checkpoint and continue the job or job step
118
119 disable Disable future checkpoints
120
121 enable Enable future checkpoints
122
123 error Report the result for the last checkpoint request,
124 error code and message
125
126 restart Restart execution of the previously checkpointed job
127 or job step
128
129 requeue Create a checkpoint and requeue the batch job, com‐
130 bines vacate and restart operations
131
132 vacate Create a checkpoint and terminate the job or job
133 step
134 Acceptable values for CKPT_OP include:
135
136 MaxWait=<seconds> Maximum time for checkpoint to be written.
137 Default value is 10 seconds. Valid with
138 create and vacate options only.
139
140 ImageDir=<directory_name>
141 Location of checkpoint file. Valid with
142 create, vacate and restart options only.
143 This value takes precedent over any --check‐
144 point-dir value specified at job submission
145 time.
146
147 StickToNodes If set, resume job on the same nodes are
148 previously used. Valid with the restart
149 option only.
150
151
152 cluster CLUSTER_NAME
153 The cluster to issue commands to. Only one cluster name may be
154 specified.
155
156
157 create SPECIFICATION
158 Create a new partition or reservation. See the full list of
159 parameters below. Include the tag "res" to create a reservation
160 without specifying a reservation name.
161
162
163 completing
164 Display all jobs in a COMPLETING state along with associated
165 nodes in either a COMPLETING or DOWN state.
166
167
168 delete SPECIFICATION
169 Delete the entry with the specified SPECIFICATION. The two
170 SPECIFICATION choices are PartitionName=<name> and Reserva‐
171 tion=<name>. Reservations and partitions should have no associ‐
172 ated jobs at the time of their deletion (modify the jobs first).
173 If the specified partition is in use, the request is denied.
174
175
176 errnumstr ERRNO
177 Given a Slurm error number, return a descriptive string.
178
179
180 fsdampeningfactor FACTOR
181 Set the FairShareDampeningFactor in slurmctld.
182
183
184 help Display a description of scontrol options and commands.
185
186
187 hold job_list
188 Prevent a pending job from being started (sets it's priority to
189 0). Use the release command to permit the job to be scheduled.
190 The job_list argument is a comma separated list of job IDs OR
191 "jobname=" with the job's name, which will attempt to hold all
192 jobs having that name. Note that when a job is held by a system
193 administrator using the hold command, only a system administra‐
194 tor may release the job for execution (also see the uhold com‐
195 mand). When the job is held by its owner, it may also be
196 released by the job's owner. Additionally, attempting to hold a
197 running job will have not suspend or cancel it. But, it will set
198 the job priority to 0 and update the job reason field, which
199 would hold the job if it was requeued at a later time.
200
201
202 notify job_id message
203 Send a message to standard error of the salloc or srun command
204 or batch job associated with the specified job_id.
205
206
207 pidinfo proc_id
208 Print the Slurm job id and scheduled termination time corre‐
209 sponding to the supplied process id, proc_id, on the current
210 node. This will work only with processes on node on which scon‐
211 trol is run, and only for those processes spawned by Slurm and
212 their descendants.
213
214
215 listpids [job_id[.step_id]] [NodeName]
216 Print a listing of the process IDs in a job step (if
217 JOBID.STEPID is provided), or all of the job steps in a job (if
218 job_id is provided), or all of the job steps in all of the jobs
219 on the local node (if job_id is not provided or job_id is "*").
220 This will work only with processes on the node on which scontrol
221 is run, and only for those processes spawned by Slurm and their
222 descendants. Note that some Slurm configurations (ProctrackType
223 value of pgid) are unable to identify all processes associated
224 with a job or job step.
225
226 Note that the NodeName option is only really useful when you
227 have multiple slurmd daemons running on the same host machine.
228 Multiple slurmd daemons on one host are, in general, only used
229 by Slurm developers.
230
231
232 ping Ping the primary and secondary slurmctld daemon and report if
233 they are responding.
234
235
236 reboot [ASAP] [nextstate=<RESUME|DOWN>] [reason=<reason>] [NodeList]
237 Reboot all nodes in the system when they become idle using the
238 RebootProgram as configured in Slurm's slurm.conf file. Each
239 node will have the "REBOOT" flag added to its node state. After
240 a node reboots and the slurmd daemon starts up again, the
241 HealthCheckProgram will run once. Then, the slurmd daemon will
242 register itself with the slurmctld daemon and the "REBOOT" flag
243 will be cleared. If the node reason is "Reboot ASAP", Slurm
244 will clear the node's "DRAIN" state flag as well. The "ASAP"
245 option adds the "DRAIN" flag to each node's state, preventing
246 additional jobs from running on the node so it can be rebooted
247 and returned to service "As Soon As Possible" (i.e. ASAP).
248 "ASAP" will also set the node reason to "Reboot ASAP" if the
249 "reason" option isn't specified. If the "nextstate" option is
250 specified as "DOWN", then the node will remain in a down state
251 after rebooting. If "nextstate" is specified as "RESUME", then
252 the nodes will resume as normal when the node registers and the
253 node reason will be cleared. Resuming nodes will be considered
254 as available in backfill future scheduling and won't be replaced
255 by idle nodes in a reservation. When using the "nextstate" and
256 "reason" options together the reason will be appended with
257 "reboot issued" when the reboot is issued and "reboot complete"
258 when the node registers with a "nextstate" of "DOWN". The "rea‐
259 son" option sets each node's reason to a user-defined message.
260 An optional list of nodes to reboot may be specified. By default
261 all nodes are rebooted. NOTE: By default, this command does not
262 prevent additional jobs from being scheduled on any nodes before
263 reboot. To do this, you can either use the "ASAP" option or
264 explicitly drain the nodes beforehand. You can alternately cre‐
265 ate an advanced reservation to prevent additional jobs from
266 being initiated on nodes to be rebooted. Pending reboots can be
267 cancelled by using "scontrol cancel_reboot <node>" or setting
268 the node state to "CANCEL_REBOOT". A node will be marked "DOWN"
269 if it doesn't reboot within ResumeTimeout.
270
271
272 reconfigure
273 Instruct all Slurm daemons to re-read the configuration file.
274 This command does not restart the daemons. This mechanism would
275 be used to modify configuration parameters (Epilog, Prolog,
276 SlurmctldLogFile, SlurmdLogFile, etc.). The Slurm controller
277 (slurmctld) forwards the request all other daemons (slurmd dae‐
278 mon on each compute node). Running jobs continue execution.
279 Most configuration parameters can be changed by just running
280 this command, however, Slurm daemons should be shutdown and
281 restarted if any of these parameters are to be changed:
282 AuthType, BackupAddr, BackupController, ControlAddr, Control‐
283 Mach, PluginDir, StateSaveLocation, SlurmctldPort or SlurmdPort.
284 The slurmctld daemon and all slurmd daemons must be restarted if
285 nodes are added to or removed from the cluster.
286
287
288 release job_list
289 Release a previously held job to begin execution. The job_list
290 argument is a comma separated list of job IDs OR "jobname=" with
291 the job's name, which will attempt to hold all jobs having that
292 name. Also see hold.
293
294
295 requeue job_list
296 Requeue a running, suspended or finished Slurm batch job into
297 pending state. The job_list argument is a comma separated list
298 of job IDs.
299
300
301 requeuehold job_list
302 Requeue a running, suspended or finished Slurm batch job into
303 pending state, moreover the job is put in held state (priority
304 zero). The job_list argument is a comma separated list of job
305 IDs. A held job can be released using scontrol to reset its
306 priority (e.g. "scontrol release <job_id>"). The command
307 accepts the following option:
308
309 State=SpecialExit
310 The "SpecialExit" keyword specifies that the job has to
311 be put in a special state JOB_SPECIAL_EXIT. The "scon‐
312 trol show job" command will display the JobState as SPE‐
313 CIAL_EXIT, while the "squeue" command as SE.
314
315
316 resume job_list
317 Resume a previously suspended job. The job_list argument is a
318 comma separated list of job IDs. Also see suspend.
319
320 NOTE: A suspended job releases its CPUs for allocation to other
321 jobs. Resuming a previously suspended job may result in multi‐
322 ple jobs being allocated the same CPUs, which could trigger gang
323 scheduling with some configurations or severe degradation in
324 performance with other configurations. Use of the scancel com‐
325 mand to send SIGSTOP and SIGCONT signals would stop a job with‐
326 out releasing its CPUs for allocation to other jobs and would be
327 a preferable mechanism in many cases. If performing system
328 maintenance you may want to use suspend/resume in the following
329 way. Before suspending set all nodes to draining or set all par‐
330 titions to down so that no new jobs can be scheduled. Then sus‐
331 pend jobs. Once maintenance is done resume jobs then resume
332 nodes and/or set all partitions back to up. Use with caution.
333
334
335 schedloglevel LEVEL
336 Enable or disable scheduler logging. LEVEL may be "0", "1",
337 "disable" or "enable". "0" has the same effect as "disable". "1"
338 has the same effect as "enable". This value is temporary and
339 will be overwritten when the slurmctld daemon reads the
340 slurm.conf configuration file (e.g. when the daemon is restarted
341 or scontrol reconfigure is executed) if the SlurmSchedLogLevel
342 parameter is present.
343
344
345 setdebug LEVEL
346 Change the debug level of the slurmctld daemon. LEVEL may be an
347 integer value between zero and nine (using the same values as
348 SlurmctldDebug in the slurm.conf file) or the name of the most
349 detailed message type to be printed: "quiet", "fatal", "error",
350 "info", "verbose", "debug", "debug2", "debug3", "debug4", or
351 "debug5". This value is temporary and will be overwritten when‐
352 ever the slurmctld daemon reads the slurm.conf configuration
353 file (e.g. when the daemon is restarted or scontrol reconfigure
354 is executed).
355
356
357 setdebugflags [+|-]FLAG
358 Add or remove DebugFlags of the slurmctld daemon. See "man
359 slurm.conf" for a list of supported DebugFlags. NOTE: Changing
360 the value of some DebugFlags will have no effect without
361 restarting the slurmctld daemon, which would set DebugFlags
362 based upon the contents of the slurm.conf configuration file.
363
364
365 show ENTITY ID
366 or
367
368 show ENTITY=ID
369 Display the state of the specified entity with the specified
370 identification. ENTITY may be aliases, assoc_mgr, bbstat,
371 burstbuffer, config, daemons, dwstat, federation, frontend, job,
372 node, partition, powercap, reservation, slurmd, step, topology,
373 hostlist, hostlistsorted or hostnames ID can be used to identify
374 a specific element of the identified entity: job ID, node name,
375 partition name, reservation name, or job step ID for job, node,
376 partition, or step respectively. For an ENTITY of bbstat or
377 dwstat (they are equivalent) optional arguments are the options
378 of the local status command. The status commands will be exe‐
379 cuted by the slurmctld daemon and its response returned to the
380 user. For an ENTITY of topology, the ID may be a node or switch
381 name. If one node name is specified, all switches connected to
382 that node (and their parent switches) will be shown. If more
383 than one node name is specified, only switches that connect to
384 all named nodes will be shown. aliases will return all NodeName
385 values associated to a given NodeHostname (useful to get the
386 list of virtual nodes associated with a real node in a configu‐
387 ration where multiple slurmd daemons execute on a single compute
388 node). assoc_mgr displays the current contents of the slurm‐
389 ctld's internal cache for users, associations and/or qos. The ID
390 may be users=<user1>,[...,<userN>],
391 accounts=<acct1>,[...,<acctN>], qos=<qos1>,[...,<qosN>] and/or
392 flags=<users,assoc,qos>, used to filter the desired section to
393 be displayed. If no flags are specified, all sections are dis‐
394 played. burstbuffer displays the current status of the Burst‐
395 Buffer plugin. config displays parameter names from the config‐
396 uration files in mixed case (e.g. SlurmdPort=7003) while derived
397 parameters names are in upper case only (e.g. SLURM_VERSION).
398 hostnames takes an optional hostlist expression as input and
399 writes a list of individual host names to standard output (one
400 per line). If no hostlist expression is supplied, the contents
401 of the SLURM_JOB_NODELIST environment variable is used. For
402 example "tux[1-3]" is mapped to "tux1","tux2" and "tux3" (one
403 hostname per line). hostlist takes a list of host names and
404 prints the hostlist expression for them (the inverse of host‐
405 names). hostlist can also take the absolute pathname of a file
406 (beginning with the character '/') containing a list of host‐
407 names. Multiple node names may be specified using simple node
408 range expressions (e.g. "lx[10-20]"). All other ID values must
409 identify a single element. The job step ID is of the form
410 "job_id.step_id", (e.g. "1234.1"). slurmd reports the current
411 status of the slurmd daemon executing on the same node from
412 which the scontrol command is executed (the local host). It can
413 be useful to diagnose problems. By default hostlist does not
414 sort the node list or make it unique (e.g. tux2,tux1,tux2 =
415 tux[2,1-2]). If you wanted a sorted list use hostlistsorted
416 (e.g. tux2,tux1,tux2 = tux[1-2,2]). By default, all elements of
417 the entity type specified are printed. For an ENTITY of job, if
418 the job does not specify socket-per-node, cores-per-socket or
419 threads-per-core then it will display '*' in ReqS:C:T=*:*:*
420 field. For an ENTITY of federation, the federation name that the
421 controller is part of and the sibling clusters part of the fed‐
422 eration will be listed.
423
424
425 shutdown OPTION
426 Instruct Slurm daemons to save current state and terminate. By
427 default, the Slurm controller (slurmctld) forwards the request
428 all other daemons (slurmd daemon on each compute node). An
429 OPTION of slurmctld or controller results in only the slurmctld
430 daemon being shutdown and the slurmd daemons remaining active.
431
432
433 suspend job_list
434 Suspend a running job. The job_list argument is a comma sepa‐
435 rated list of job IDs. Use the resume command to resume its
436 execution. User processes must stop on receipt of SIGSTOP sig‐
437 nal and resume upon receipt of SIGCONT for this operation to be
438 effective. Not all architectures and configurations support job
439 suspension. If a suspended job is requeued, it will be placed
440 in a held state. The time a job is suspended will not count
441 against a job's time limit. Only an operator, administrator,
442 SlurmUser, or root can suspend jobs.
443
444
445 takeover
446 Instruct Slurm's backup controller (slurmctld) to take over sys‐
447 tem control. Slurm's backup controller requests control from
448 the primary and waits for its termination. After that, it
449 switches from backup mode to controller mode. If primary con‐
450 troller can not be contacted, it directly switches to controller
451 mode. This can be used to speed up the Slurm controller
452 fail-over mechanism when the primary node is down. This can be
453 used to minimize disruption if the computer executing the pri‐
454 mary Slurm controller is scheduled down. (Note: Slurm's primary
455 controller will take the control back at startup.)
456
457
458 top job_list
459 Move the specified job IDs to the top of the queue of jobs
460 belonging to the identical user ID, partition name, account, and
461 QOS. The job_list argument is a comma separated ordered list of
462 job IDs. Any job not matching all of those fields will not be
463 effected. Only jobs submitted to a single partition will be
464 effected. This operation changes the order of jobs by adjusting
465 job nice values. The net effect on that user's throughput will
466 be negligible to slightly negative. This operation is disabled
467 by default for non-privileged (non-operator, admin, SlurmUser,
468 or root) users. This operation may be enabled for non-privileged
469 users by the system administrator by including the option
470 "enable_user_top" in the SchedulerParameters configuration
471 parameter.
472
473
474 uhold job_list
475 Prevent a pending job from being started (sets it's priority to
476 0). The job_list argument is a space separated list of job IDs
477 or job names. Use the release command to permit the job to be
478 scheduled. This command is designed for a system administrator
479 to hold a job so that the job owner may release it rather than
480 requiring the intervention of a system administrator (also see
481 the hold command).
482
483
484 update SPECIFICATION
485 Update job, step, node, partition, powercapping or reservation
486 configuration per the supplied specification. SPECIFICATION is
487 in the same format as the Slurm configuration file and the out‐
488 put of the show command described above. It may be desirable to
489 execute the show command (described above) on the specific
490 entity you want to update, then use cut-and-paste tools to enter
491 updated configuration values to the update. Note that while most
492 configuration values can be changed using this command, not all
493 can be changed using this mechanism. In particular, the hardware
494 configuration of a node or the physical addition or removal of
495 nodes from the cluster may only be accomplished through editing
496 the Slurm configuration file and executing the reconfigure com‐
497 mand (described above).
498
499
500 version
501 Display the version number of scontrol being executed.
502
503
504 wait_job job_id
505 Wait until a job and all of its nodes are ready for use or the
506 job has entered some termination state. This option is particu‐
507 larly useful in the Slurm Prolog or in the batch script itself
508 if nodes are powered down and restarted automatically as needed.
509
510
511 write batch_script job_id optional_filename
512 Write the batch script for a given job_id to a file or to std‐
513 out. The file will default to slurm-<job_id>.sh if the optional
514 filename argument is not given. The script will be written to
515 stdout if - is given instead of a filename. The batch script
516 can only be retrieved by an admin or operator, or by the owner
517 of the job.
518
519
520 write config
521 Write the current configuration to a file with the naming con‐
522 vention of "slurm.conf.<datetime>" in the same directory as the
523 original slurm.conf file.
524
525
526 INTERACTIVE COMMANDS
527 NOTE: All commands listed below can be used in the interactive
528 mode, but NOT on the initial command line.
529
530
531 all Show all partitions, their jobs and jobs steps. This causes
532 information to be displayed about partitions that are configured
533 as hidden and partitions that are unavailable to user's group.
534
535
536 details
537 Causes the show command to provide additional details where
538 available. Job information will include CPUs and NUMA memory
539 allocated on each node. Note that on computers with hyper‐
540 threading enabled and Slurm configured to allocate cores, each
541 listed CPU represents one physical core. Each hyperthread on
542 that core can be allocated a separate task, so a job's CPU count
543 and task count may differ. See the --cpu-bind and --mem-bind
544 option descriptions in srun man pages for more information. The
545 details option is currently only supported for the show job com‐
546 mand.
547
548
549 exit Terminate scontrol interactive session.
550
551
552 hide Do not display partition, job or jobs step information for par‐
553 titions that are configured as hidden or partitions that are
554 unavailable to the user's group. This is the default behavior.
555
556
557 oneliner
558 Print information one line per record.
559
560
561 quiet Print no warning or informational messages, only fatal error
562 messages.
563
564
565 quit Terminate the execution of scontrol.
566
567
568 verbose
569 Print detailed event logging. This includes time-stamps on data
570 structures, record counts, etc.
571
572
573 !! Repeat the last command executed.
574
575
576 SPECIFICATIONS FOR UPDATE COMMAND, JOBS
577
578 Note that update requests done by either root, SlurmUser or Administra‐
579 tors are not subject to certain restrictions. For instance, if an
580 Administrator changes the QOS on a pending job, certain limits such as
581 the TimeLimit will not be changed automatically as changes made by the
582 Administrators are allowed to violate these restrictions.
583
584
585 Account=<account>
586 Account name to be changed for this job's resource use. Value
587 may be cleared with blank data value, "Account=".
588
589 AdminComment=<spec>
590 Arbitrary descriptive string. Can only be set by a Slurm admin‐
591 istrator.
592
593 ArrayTaskThrottle=<count>
594 Specify the maximum number of tasks in a job array that can exe‐
595 cute at the same time. Set the count to zero in order to elimi‐
596 nate any limit. The task throttle count for a job array is
597 reported as part of its ArrayTaskId field, preceded with a per‐
598 cent sign. For example "ArrayTaskId=1-10%2" indicates the maxi‐
599 mum number of running tasks is limited to 2.
600
601 BurstBuffer=<spec>
602 Burst buffer specification to be changed for this job's resource
603 use. Value may be cleared with blank data value, "Burst‐
604 Buffer=". Format is burst buffer plugin specific.
605
606 Clusters=<spec>
607 Specifies the clusters that the federated job can run on.
608
609 ClusterFeatures=<spec>
610 Specifies features that a federated cluster must have to have a
611 sibling job submitted to it. Slurm will attempt to submit a sib‐
612 ling job to a cluster if it has at least one of the specified
613 features.
614
615 Comment=<spec>
616 Arbitrary descriptive string.
617
618 Contiguous=<yes|no>
619 Set the job's requirement for contiguous (consecutive) nodes to
620 be allocated. Possible values are "YES" and "NO". Only the
621 Slurm administrator or root can change this parameter.
622
623 CoreSpec=<count>
624 Number of cores to reserve per node for system use. The job
625 will be charged for these cores, but be unable to use them.
626 Will be reported as "*" if not constrained.
627
628 CPUsPerTask=<count>
629 Change the CPUsPerTask job's value.
630
631 Deadline=<time_spec>
632 It accepts times of the form HH:MM:SS to specify a deadline to a
633 job at a specific time of day (seconds are optional). You may
634 also specify midnight, noon, fika (3 PM) or teatime (4 PM) and
635 you can have a time-of-day suffixed with AM or PM for a deadline
636 in the morning or the evening. You can specify a deadline for
637 the job with a date of the form MMDDYY or MM/DD/YY or MM.DD.YY,
638 or a date and time as YYYY-MM-DD[THH:MM[:SS]]. You can also
639 give times like now + count time-units, where the time-units can
640 be minutes, hours, days, or weeks and you can tell Slurm to put
641 a deadline for tomorrow with the keyword tomorrow. The speci‐
642 fied deadline must be later than the current time. Only pending
643 jobs can have the deadline updated. Only the Slurm administra‐
644 tor or root can change this parameter.
645
646 DelayBoot=<time_spec>
647 Change the time to decide whether to reboot nodes in order to
648 satisfy job's feature specification if the job has been eligible
649 to run for less than this time period. See salloc/sbatch man
650 pages option --delay-boot.
651
652 Dependency=<dependency_list>
653 Defer job's initiation until specified job dependency specifica‐
654 tion is satisfied. Cancel dependency with an empty depen‐
655 dency_list (e.g. "Dependency="). <dependency_list> is of the
656 form <type:job_id[:job_id][,type:job_id[:job_id]]>. Many jobs
657 can share the same dependency and these jobs may even belong to
658 different users.
659
660 after:job_id[:jobid...]
661 This job can begin execution after the specified jobs
662 have begun execution.
663
664 afterany:job_id[:jobid...]
665 This job can begin execution after the specified jobs
666 have terminated.
667
668 afternotok:job_id[:jobid...]
669 This job can begin execution after the specified jobs
670 have terminated in some failed state (non-zero exit code,
671 node failure, timed out, etc).
672
673 afterok:job_id[:jobid...]
674 This job can begin execution after the specified jobs
675 have successfully executed (ran to completion with an
676 exit code of zero).
677
678 singleton
679 This job can begin execution after any previously
680 launched jobs sharing the same job name and user have
681 terminated. In other words, only one job by that name
682 and owned by that user can be running or suspended at any
683 point in time.
684
685 EligibleTime=<time_spec>
686 See StartTime.
687
688 EndTime
689 The time the job is expected to terminate based on the job's
690 time limit. When the job ends sooner, this field will be
691 updated with the actual end time.
692
693 ExcNodeList=<nodes>
694 Set the job's list of excluded node. Multiple node names may be
695 specified using simple node range expressions (e.g.
696 "lx[10-20]"). Value may be cleared with blank data value, "Exc‐
697 NodeList=".
698
699 Features=<features>
700 Set the job's required node features. The list of features may
701 include multiple feature names separated by ampersand (AND)
702 and/or vertical bar (OR) operators. For example: Fea‐
703 tures="opteron&video" or Features="fast|faster". In the first
704 example, only nodes having both the feature "opteron" AND the
705 feature "video" will be used. There is no mechanism to specify
706 that you want one node with feature "opteron" and another node
707 with feature "video" in case no node has both features. If only
708 one of a set of possible options should be used for all allo‐
709 cated nodes, then use the OR operator and enclose the options
710 within square brackets. For example: "Fea‐
711 tures=[rack1|rack2|rack3|rack4]" might be used to specify that
712 all nodes must be allocated on a single rack of the cluster, but
713 any of those four racks can be used. A request can also specify
714 the number of nodes needed with some feature by appending an
715 asterisk and count after the feature name. For example "Fea‐
716 tures=graphics*4" indicates that at least four allocated nodes
717 must have the feature "graphics." Parenthesis are also sup‐
718 ported for features to be ANDed together. For example "Fea‐
719 tures=[(knl&a2a&flat)*4&haswell*2]" indicates the resource allo‐
720 cation should include 4 nodes with ALL of the features "knl",
721 "a2a", and "flat" plus 2 nodes with the feature "haswell". Con‐
722 straints with node counts may only be combined with AND opera‐
723 tors. Value may be cleared with blank data value, for example
724 "Features=".
725
726
727 Gres=<list>
728 Specifies a comma delimited list of generic consumable
729 resources. The format of each entry on the list is
730 "name[:count[*cpu]]". The name is that of the consumable
731 resource. The count is the number of those resources with a
732 default value of 1. The specified resources will be allocated
733 to the job on each node allocated unless "*cpu" is appended, in
734 which case the resources will be allocated on a per cpu basis.
735 The available generic consumable resources is configurable by
736 the system administrator. A list of available generic consum‐
737 able resources will be printed and the command will exit if the
738 option argument is "help". Examples of use include
739 "Gres=gpus:2*cpu,disk=40G" and "Gres=help".
740
741
742 JobId=<job_list>
743 Identify the job(s) to be updated. The job_list may be a comma
744 separated list of job IDs. Either JobId or JobName is required.
745
746 Licenses=<name>
747 Specification of licenses (or other resources available on all
748 nodes of the cluster) as described in salloc/sbatch/srun man
749 pages.
750
751 MinCPUsNode=<count>
752 Set the job's minimum number of CPUs per node to the specified
753 value.
754
755 MinMemoryCPU=<megabytes>
756 Set the job's minimum real memory required per allocated CPU to
757 the specified value. Either MinMemoryCPU or MinMemoryNode may be
758 set, but not both.
759
760 MinMemoryNode=<megabytes>
761 Set the job's minimum real memory required per node to the spec‐
762 ified value. Either MinMemoryCPU or MinMemoryNode may be set,
763 but not both.
764
765 MinTmpDiskNode=<megabytes>
766 Set the job's minimum temporary disk space required per node to
767 the specified value. Only the Slurm administrator or root can
768 change this parameter.
769
770 imeMin=<timespec>
771 Change TimeMin value which specifies the minimum time limit min‐
772 utes of the job.
773
774 JobName=<name>
775 Identify the name of jobs to be modified or set the job's name
776 to the specified value. When used to identify jobs to be modi‐
777 fied, all jobs belonging to all users are modified unless the
778 UserID option is used to identify a specific user. Either JobId
779 or JobName is required.
780
781 Name[=<name>]
782 See JobName.
783
784 Nice[=<adjustment>]
785 Update the job with an adjusted scheduling priority within
786 Slurm. With no adjustment value the scheduling priority is
787 decreased by 100. A negative nice value increases the priority,
788 otherwise decreases it. The adjustment range is +/- 2147483645.
789 Only privileged users can specify a negative adjustment.
790
791 NodeList=<nodes>
792 Change the nodes allocated to a running job to shrink it's size.
793 The specified list of nodes must be a subset of the nodes cur‐
794 rently allocated to the job. Multiple node names may be speci‐
795 fied using simple node range expressions (e.g. "lx[10-20]").
796 After a job's allocation is reduced, subsequent srun commands
797 must explicitly specify node and task counts which are valid for
798 the new allocation.
799
800 NumCPUs=<min_count>[-<max_count>]
801 Set the job's minimum and optionally maximum count of CPUs to be
802 allocated.
803
804 NumNodes=<min_count>[-<max_count>]
805 Set the job's minimum and optionally maximum count of nodes to
806 be allocated. If the job is already running, use this to spec‐
807 ify a node count less than currently allocated and resources
808 previously allocated to the job will be relinquished. After a
809 job's allocation is reduced, subsequent srun commands must
810 explicitly specify node and task counts which are valid for the
811 new allocation. Also see the NodeList parameter above. This is
812 the same than ReqNodes.
813
814 NumTasks=<count>
815 Set the job's count of required tasks to the specified value.
816 This is the same than ReqProcs.
817
818 OverSubscribe=<yes|no>
819 Set the job's ability to share compute resources (i.e. individ‐
820 ual CPUs) with other jobs. Possible values are "YES" and "NO".
821 This option can only be changed for pending jobs.
822
823 Partition=<name>
824 Set the job's partition to the specified value.
825
826 Priority=<number>
827 Set the job's priority to the specified value. Note that a job
828 priority of zero prevents the job from ever being scheduled. By
829 setting a job's priority to zero it is held. Set the priority
830 to a non-zero value to permit it to run. Explicitly setting a
831 job's priority clears any previously set nice value and removes
832 the priority/multifactor plugin's ability to manage a job's pri‐
833 ority. In order to restore the priority/multifactor plugin's
834 ability to manage a job's priority, hold and then release the
835 job. Only the Slurm administrator or root can increase job's
836 priority.
837
838 QOS=<name>
839 Set the job's QOS (Quality Of Service) to the specified value.
840 Value may be cleared with blank data value, "QOS=".
841
842 Reboot=<yes|no>
843 Set the job's flag that specifies whether to force the allocated
844 nodes to reboot before starting the job. This is only supported
845 with some system configurations and therefore it could be
846 silently ignored.
847
848 ReqCores=<count>
849 Change the job's requested Cores count.
850
851 ReqNodeList=<nodes>
852 Set the job's list of required node. Multiple node names may be
853 specified using simple node range expressions (e.g.
854 "lx[10-20]"). Value may be cleared with blank data value,
855 "ReqNodeList=".
856
857 ReqNodes=<min_count>[-<max_count>]
858 See NumNodes.
859
860 ReqProcs=<count>
861 See NumTasks.
862
863 ReqSockets=<count>
864 Change the job's requested socket count.
865
866 ReqThreads=<count>
867 Change the job's requested threads count.
868
869 Requeue=<0|1>
870 Stipulates whether a job should be requeued after a node fail‐
871 ure: 0 for no, 1 for yes.
872
873 ReservationName=<name>
874 Set the job's reservation to the specified value. Value may be
875 cleared with blank data value, "ReservationName=".
876
877 ResetAccrueTime
878 Reset the job's accrue time value to 0 meaning it will loose any
879 time previously accrued for priority. Helpful if you have a
880 large queue of jobs already in the queue and want to start lim‐
881 iting how many jobs can accrue time without waiting for the
882 queue to flush out.
883
884 StdOut=<filepath>
885 Set the batch job's stdout file path.
886
887 Shared=<yes|no>
888 See OverSubscribe option above.
889
890 StartTime=<time_spec>
891 Set the job's earliest initiation time. It accepts times of the
892 form HH:MM:SS to run a job at a specific time of day (seconds
893 are optional). (If that time is already past, the next day is
894 assumed.) You may also specify midnight, noon, fika (3 PM) or
895 teatime (4 PM) and you can have a time-of-day suffixed with AM
896 or PM for running in the morning or the evening. You can also
897 say what day the job will be run, by specifying a date of the
898 form MMDDYY or MM/DD/YY or MM.DD.YY, or a date and time as
899 YYYY-MM-DD[THH:MM[:SS]]. You can also give times like now +
900 count time-units, where the time-units can be minutes, hours,
901 days, or weeks and you can tell Slurm to run the job today with
902 the keyword today and to run the job tomorrow with the keyword
903 tomorrow.
904
905 Notes on date/time specifications:
906 - although the 'seconds' field of the HH:MM:SS time specifica‐
907 tion is allowed by the code, note that the poll time of the
908 Slurm scheduler is not precise enough to guarantee dispatch of
909 the job on the exact second. The job will be eligible to start
910 on the next poll following the specified time. The exact poll
911 interval depends on the Slurm scheduler (e.g., 60 seconds with
912 the default sched/builtin).
913 - if no time (HH:MM:SS) is specified, the default is
914 (00:00:00).
915 - if a date is specified without a year (e.g., MM/DD) then the
916 current year is assumed, unless the combination of MM/DD and
917 HH:MM:SS has already passed for that year, in which case the
918 next year is used.
919
920 Switches=<count>[@<max-time-to-wait>]
921 When a tree topology is used, this defines the maximum count of
922 switches desired for the job allocation. If Slurm finds an allo‐
923 cation containing more switches than the count specified, the
924 job remain pending until it either finds an allocation with
925 desired switch count or the time limit expires. By default there
926 is no switch count limit and no time limit delay. Set the count
927 to zero in order to clean any previously set count (disabling
928 the limit). The job's maximum time delay may be limited by the
929 system administrator using the SchedulerParameters configuration
930 parameter with the max_switch_wait parameter option. Also see
931 wait-for-switch.
932
933
934 wait-for-switch=<seconds>
935 Change max time to wait for a switch <seconds> secs.
936
937
938 TasksPerNode=<count>
939 Change the job's requested TasksPerNode.
940
941
942 ThreadSpec=<count>
943 Number of threads to reserve per node for system use. The job
944 will be charged for these threads, but be unable to use them.
945 Will be reported as "*" if not constrained.
946
947
948 TimeLimit=<time>
949 The job's time limit. Output format is [days-]hours:min‐
950 utes:seconds or "UNLIMITED". Input format (for update command)
951 set is minutes, minutes:seconds, hours:minutes:seconds,
952 days-hours, days-hours:minutes or days-hours:minutes:seconds.
953 Time resolution is one minute and second values are rounded up
954 to the next minute. If changing the time limit of a job, either
955 specify a new time limit value or precede the time and equal
956 sign with a "+" or "-" to increment or decrement the current
957 time limit (e.g. "TimeLimit+=30"). In order to increment or
958 decrement the current time limit, the JobId specification must
959 precede the TimeLimit specification. Only the Slurm administra‐
960 tor or root can increase job's TimeLimit.
961
962
963 UserID=<UID or name>
964 Used with the JobName option to identify jobs to be modified.
965 Either a user name or numeric ID (UID), may be specified.
966
967
968 WCKey=<key>
969 Set the job's workload characterization key to the specified
970 value.
971
972
973 SPECIFICATIONS FOR SHOW COMMAND, JOBS
974
975 The "show" command, when used with the "job" or "job <jobid>" entity
976 displays detailed information about a job or jobs. Much of this infor‐
977 mation may be modified using the "update job" command as described
978 above. However, the following fields displayed by the show job command
979 are read-only and cannot be modified:
980
981
982 AllocNode:Sid
983 Local node and system id making the resource allocation.
984
985 BatchFlag
986 Jobs submitted using the sbatch command have BatchFlag set to 1.
987 Jobs submitted using other commands have BatchFlag set to 0.
988
989 ExitCode=<exit>:<sig>
990 Exit status reported for the job by the wait() function. The
991 first number is the exit code, typically as set by the exit()
992 function. The second number of the signal that caused the
993 process to terminate if it was terminated by a signal.
994
995 GroupId
996 The group under which the job was submitted.
997
998 JobState
999 The current state of the job.
1000
1001 NodeListIndices
1002 The NodeIndices expose the internal indices into the node table
1003 associated with the node(s) allocated to the job.
1004
1005 NtasksPerN:B:S:C=
1006 <tasks_per_node>:<tasks_per_base‐
1007 board>:<tasks_per_socket>:<tasks_per_core> Specifies the number
1008 of tasks to be started per hardware component (node, baseboard,
1009 socket and core). Unconstrained values may be shown as "0" or
1010 "*".
1011
1012 PreemptTime
1013 Time at which job was signaled that it was selected for preemp‐
1014 tion. (Meaningful only for PreemptMode=CANCEL and the partition
1015 or QOS with which the job is associated has a GraceTime value
1016 designated.)
1017
1018 PreSusTime
1019 Time the job ran prior to last suspend.
1020
1021 Reason The reason job is not running: e.g., waiting "Resources".
1022
1023 ReqB:S:C:T=
1024 <baseboard_count>:<socket_per_base‐
1025 board_count>:<core_per_socket_count>:<thread_per_core_count>
1026 Specifies the count of various hardware components requested by
1027 the job. Unconstrained values may be shown as "0" or "*".
1028
1029 SecsPreSuspend=<seconds>
1030 If the job is suspended, this is the run time accumulated by the
1031 job (in seconds) prior to being suspended.
1032
1033 Socks/Node=<count>
1034 Count of desired sockets per node
1035
1036 SubmitTime
1037 The time and date stamp (in localtime) the job was submitted.
1038 The format of the output is identical to that of the EndTime
1039 field.
1040
1041 NOTE: If a job is requeued, the submit time is reset. To obtain
1042 the original submit time it is necessary to use the "sacct -j
1043 <job_id[.<step_id>]" command also designating the -D or --dupli‐
1044 cate option to display all duplicate entries for a job.
1045
1046 SuspendTime
1047 Time the job was last suspended or resumed.
1048
1049 NOTE on information displayed for various job states:
1050 When you submit a request for the "show job" function the scon‐
1051 trol process makes an RPC request call to slurmctld with a
1052 REQUEST_JOB_INFO message type. If the state of the job is PEND‐
1053 ING, then it returns some detail information such as: min_nodes,
1054 min_procs, cpus_per_task, etc. If the state is other than PEND‐
1055 ING the code assumes that it is in a further state such as RUN‐
1056 NING, COMPLETE, etc. In these cases the code explicitly returns
1057 zero for these values. These values are meaningless once the job
1058 resources have been allocated and the job has started.
1059
1060
1061 SPECIFICATIONS FOR UPDATE COMMAND, STEPS
1062
1063 StepId=<job_id>[.<step_id>]
1064 Identify the step to be updated. If the job_id is given, but no
1065 step_id is specified then all steps of the identified job will
1066 be modified. This specification is required.
1067
1068 CompFile=<completion file>
1069 Update a step with information about a steps completion. Can be
1070 useful if step statistics aren't directly available through a
1071 jobacct_gather plugin. The file is a space-delimited file with
1072 format for Version 1 is as follows
1073
1074 1 34461 0 2 0 3 1361906011 1361906015 1 1 3368 13357 /bin/sleep
1075 A B C D E F G H I J K L M
1076
1077 Field Descriptions:
1078
1079 A file version
1080 B ALPS apid
1081 C inblocks
1082 D outblocks
1083 E exit status
1084 F number of allocated CPUs
1085 G start time
1086 H end time
1087 I utime
1088 J stime
1089 K maxrss
1090 L uid
1091 M command name
1092
1093 TimeLimit=<time>
1094 The job's time limit. Output format is [days-]hours:min‐
1095 utes:seconds or "UNLIMITED". Input format (for update command)
1096 set is minutes, minutes:seconds, hours:minutes:seconds,
1097 days-hours, days-hours:minutes or days-hours:minutes:seconds.
1098 Time resolution is one minute and second values are rounded up
1099 to the next minute. If changing the time limit of a step,
1100 either specify a new time limit value or precede the time with a
1101 "+" or "-" to increment or decrement the current time limit
1102 (e.g. "TimeLimit=+30"). In order to increment or decrement the
1103 current time limit, the StepId specification must precede the
1104 TimeLimit specification.
1105
1106
1107 SPECIFICATIONS FOR UPDATE COMMAND, NODES
1108
1109 NodeName=<name>
1110 Identify the node(s) to be updated. Multiple node names may be
1111 specified using simple node range expressions (e.g.
1112 "lx[10-20]"). This specification is required.
1113
1114
1115 ActiveFeatures=<features>
1116 Identify the feature(s) currently active on the specified node.
1117 Any previously active feature specification will be overwritten
1118 with the new value. Also see AvailableFeatures. Typically
1119 ActiveFeatures will be identical to AvailableFeatures; however
1120 ActiveFeatures may be configured as a subset of the Available‐
1121 Features. For example, a node may be booted in multiple configu‐
1122 rations. In that case, all possible configurations may be iden‐
1123 tified as AvailableFeatures, while ActiveFeatures would identify
1124 the current node configuration.
1125
1126
1127 AvailableFeatures=<features>
1128 Identify the feature(s) available on the specified node. Any
1129 previously defined available feature specification will be over‐
1130 written with the new value. AvailableFeatures assigned via
1131 scontrol will only persist across the restart of the slurmctld
1132 daemon with the -R option and state files preserved or slurm‐
1133 ctld's receipt of a SIGHUP. Update slurm.conf with any changes
1134 meant to be persistent across normal restarts of slurmctld or
1135 the execution of scontrol reconfig. Also see ActiveFeatures.
1136
1137
1138 CpuBind=<node>
1139 Specify the task binding mode to be used by default for this
1140 node. Supported options include: "none", "board", "socket",
1141 "ldom" (NUMA), "core", "thread" and "off" (remove previous bind‐
1142 ing mode).
1143
1144
1145 Gres=<gres>
1146 Identify generic resources to be associated with the specified
1147 node. Any previously defined generic resources will be over‐
1148 written with the new value. Specifications for multiple generic
1149 resources should be comma separated. Each resource specifica‐
1150 tion consists of a name followed by an optional colon with a
1151 numeric value (default value is one) (e.g. "Gres=band‐
1152 width:10000,gpus"). Generic resources assigned via scontrol
1153 will only persist across the restart of the slurmctld daemon
1154 with the -R option and state files preserved or slurmctld's
1155 receipt of a SIGHUP. Update slurm.conf with any changes meant
1156 to be persistent across normal restarts of slurmctld or the exe‐
1157 cution of scontrol reconfig.
1158
1159
1160 Reason=<reason>
1161 Identify the reason the node is in a "DOWN", "DRAINED", "DRAIN‐
1162 ING", "FAILING" or "FAIL" state. Use quotes to enclose a reason
1163 having more than one word.
1164
1165
1166 State=<state>
1167 Identify the state to be assigned to the node. Possible node
1168 states are "NoResp", "ALLOC", "ALLOCATED", "COMPLETING", "DOWN",
1169 "DRAIN", "FAIL", "FAILING", "FUTURE" "IDLE", "MAINT", "MIXED",
1170 "PERFCTRS/NPC", "RESERVED", "POWER_DOWN", "POWER_UP", "RESUME"
1171 or "UNDRAIN". Not all of those states can be set using the scon‐
1172 trol command only the following can: "CANCEL_REBOOT", "DOWN",
1173 "DRAIN", "FAIL", "FUTURE", "RESUME", "NoResp", "POWER_DOWN",
1174 "POWER_UP" and "UNDRAIN". If a node is in a "MIXED" state it
1175 usually means the node is in multiple states. For instance if
1176 only part of the node is "ALLOCATED" and the rest of the node is
1177 "IDLE" the state will be "MIXED". If you want to remove a node
1178 from service, you typically want to set it's state to "DRAIN".
1179 "CANCEL_REBOOT" cancels a pending reboot on the node (same as
1180 "scontrol cancel_reboot <node>"). "FAILING" is similar to
1181 "DRAIN" except that some applications will seek to relinquish
1182 those nodes before the job completes. "PERFCTRS/NPC" indicates
1183 that Network Performance Counters associated with this node are
1184 in use, rendering this node as not usable for any other jobs.
1185 "RESERVED" indicates the node is in an advanced reservation and
1186 not generally available. "RESUME" is not an actual node state,
1187 but will change a node state from "DRAINED", "DRAINING", "DOWN"
1188 or "REBOOT" to either "IDLE" or "ALLOCATED" state as appropri‐
1189 ate. "UNDRAIN" clears the node from being drained (like
1190 "RESUME"), but will not change the node's base state (e.g.
1191 "DOWN"). Setting a node "DOWN" will cause all running and sus‐
1192 pended jobs on that node to be terminated. "POWER_DOWN" and
1193 "POWER_UP" will use the configured SuspendProg and ResumeProg
1194 programs to explicitly place a node in or out of a power saving
1195 mode. If a node is already in the process of being powered up or
1196 down, the command will only change the state of the node but
1197 won't have any effect until the configured ResumeTimeout or Sus‐
1198 pendTimeout is reached. Use of this command can be useful in
1199 situations where a ResumeProg like capmc in Cray machines is
1200 stalled and one wants to restore the node to "IDLE" manually, in
1201 this case rebooting the node and setting the state to
1202 "POWER_DOWN" will cancel the previous "POWER_UP" state and the
1203 node will become "IDLE". The "NoResp" state will only set the
1204 "NoResp" flag for a node without changing its underlying state.
1205 While all of the above states are valid, some of them are not
1206 valid new node states given their prior state. If the node
1207 state code printed is followed by "~", this indicates the node
1208 is presently in a power saving mode (typically running at
1209 reduced frequency). If the node state code is followed by "#",
1210 this indicates the node is presently being powered up or config‐
1211 ured. If the node state code is followed by "$", this indicates
1212 the node is currently in a reservation with a flag value of
1213 "maintenance". If the node state code is followed by "@", this
1214 indicates the node is currently scheduled to be rebooted. Gen‐
1215 erally only "DRAIN", "FAIL" and "RESUME" should be used. NOTE:
1216 The scontrol command should not be used to change node state on
1217 Cray systems. Use Cray tools such as xtprocadmin instead.
1218
1219
1220 Weight=<weight>
1221 Identify weight to be associated with specified nodes. This
1222 allows dynamic changes to weight associated with nodes, which
1223 will be used for the subsequent node allocation decisions.
1224 Weight assigned via scontrol will only persist across the
1225 restart of the slurmctld daemon with the -R option and state
1226 files preserved or slurmctld's receipt of a SIGHUP. Update
1227 slurm.conf with any changes meant to be persistent across normal
1228 restarts of slurmctld or the execution of scontrol reconfig.
1229
1230
1231 SPECIFICATIONS FOR SHOW COMMAND, NODES
1232
1233 The meaning of the energy information is as follows:
1234
1235
1236 CurrentWatts
1237 The instantaneous power consumption of the node at the time of
1238 the last node energy accounting sample, in watts.
1239
1240
1241 LowestJoules
1242 The energy consumed by the node between the last time it was
1243 powered on and the last time it was registered by slurmd, in
1244 joules.
1245
1246
1247 ConsumedJoules
1248 The energy consumed by the node between the last time it was
1249 registered by the slurmd daemon and the last node energy
1250 accounting sample, in joules.
1251
1252
1253 If the reported value is "n/s" (not supported), the node does not sup‐
1254 port the configured AcctGatherEnergyType plugin. If the reported value
1255 is zero, energy accounting for nodes is disabled.
1256
1257
1258 The meaning of the external sensors information is as follows:
1259
1260
1261 ExtSensorsJoules
1262 The energy consumed by the node between the last time it was
1263 powered on and the last external sensors plugin node sample, in
1264 joules.
1265
1266
1267
1268 ExtSensorsWatts
1269 The instantaneous power consumption of the node at the time of
1270 the last external sensors plugin node sample, in watts.
1271
1272
1273 ExtSensorsTemp
1274 The temperature of the node at the time of the last external
1275 sensors plugin node sample, in celsius.
1276
1277
1278 If the reported value is "n/s" (not supported), the node does not sup‐
1279 port the configured ExtSensorsType plugin.
1280
1281
1282 The meaning of the resource specialization information is as follows:
1283
1284
1285 CPUSpecList
1286 The list of Slurm abstract CPU IDs on this node reserved for
1287 exclusive use by the Slurm compute node daemons (slurmd, slurm‐
1288 stepd).
1289
1290
1291 MemSpecLimit
1292 The combined memory limit, in megabytes, on this node for the
1293 Slurm compute node daemons (slurmd, slurmstepd).
1294
1295
1296 The meaning of the memory information is as follows:
1297
1298
1299 RealMemory
1300 The total memory, in MB, on the node.
1301
1302
1303 AllocMem
1304 The total memory, in MB, currently allocated by jobs on the
1305 node.
1306
1307
1308 FreeMem
1309 The total memory, in MB, currently free on the node as reported
1310 by the OS.
1311
1312
1313 SPECIFICATIONS FOR UPDATE COMMAND, FRONTEND
1314
1315
1316 FrontendName=<name>
1317 Identify the front end node to be updated. This specification is
1318 required.
1319
1320
1321 Reason=<reason>
1322 Identify the reason the node is in a "DOWN" or "DRAIN" state.
1323 Use quotes to enclose a reason having more than one word.
1324
1325
1326 State=<state>
1327 Identify the state to be assigned to the front end node. Possi‐
1328 ble values are "DOWN", "DRAIN" or "RESUME". If you want to
1329 remove a front end node from service, you typically want to set
1330 it's state to "DRAIN". "RESUME" is not an actual node state,
1331 but will return a "DRAINED", "DRAINING", or "DOWN" front end
1332 node to service, either "IDLE" or "ALLOCATED" state as appropri‐
1333 ate. Setting a front end node "DOWN" will cause all running and
1334 suspended jobs on that node to be terminated.
1335
1336
1337 SPECIFICATIONS FOR CREATE, UPDATE, AND DELETE COMMANDS, PARTITIONS
1338
1339 AllowGroups=<name>
1340 Identify the user groups which may use this partition. Multiple
1341 groups may be specified in a comma separated list. To permit
1342 all groups to use the partition specify "AllowGroups=ALL".
1343
1344
1345 AllocNodes=<name>
1346 Comma separated list of nodes from which users can execute jobs
1347 in the partition. Node names may be specified using the node
1348 range expression syntax described above. The default value is
1349 "ALL".
1350
1351
1352 Alternate=<partition name>
1353 Alternate partition to be used if the state of this partition is
1354 "DRAIN" or "INACTIVE." The value "NONE" will clear a previously
1355 set alternate partition.
1356
1357
1358 CpuBind=<node>
1359 Specify the task binding mode to be used by default for this
1360 partition. Supported options include: "none", "board",
1361 "socket", "ldom" (NUMA), "core", "thread" and "off" (remove pre‐
1362 vious binding mode).
1363
1364
1365 Default=<yes|no>
1366 Specify if this partition is to be used by jobs which do not
1367 explicitly identify a partition to use. Possible output values
1368 are "YES" and "NO". In order to change the default partition of
1369 a running system, use the scontrol update command and set
1370 Default=yes for the partition that you want to become the new
1371 default.
1372
1373
1374 DefaultTime=<time>
1375 Run time limit used for jobs that don't specify a value. If not
1376 set then MaxTime will be used. Format is the same as for Max‐
1377 Time.
1378
1379
1380 DefMemPerCPU=<MB>
1381 Set the default memory to be allocated per CPU for jobs in this
1382 partition. The memory size is specified in megabytes.
1383
1384 DefMemPerNode=<MB>
1385 Set the default memory to be allocated per node for jobs in this
1386 partition. The memory size is specified in megabytes.
1387
1388
1389 DisableRootJobs=<yes|no>
1390 Specify if jobs can be executed as user root. Possible values
1391 are "YES" and "NO".
1392
1393
1394 GraceTime=<seconds>
1395 Specifies, in units of seconds, the preemption grace time to be
1396 extended to a job which has been selected for preemption. The
1397 default value is zero, no preemption grace time is allowed on
1398 this partition or qos. (Meaningful only for PreemptMode=CANCEL)
1399
1400
1401 Hidden=<yes|no>
1402 Specify if the partition and its jobs should be hidden from
1403 view. Hidden partitions will by default not be reported by
1404 Slurm APIs or commands. Possible values are "YES" and "NO".
1405
1406
1407 MaxMemPerCPU=<MB>
1408 Set the maximum memory to be allocated per CPU for jobs in this
1409 partition. The memory size is specified in megabytes.
1410
1411 MaxMemPerCNode=<MB>
1412 Set the maximum memory to be allocated per node for jobs in this
1413 partition. The memory size is specified in megabytes.
1414
1415
1416 MaxNodes=<count>
1417 Set the maximum number of nodes which will be allocated to any
1418 single job in the partition. Specify a number, "INFINITE" or
1419 "UNLIMITED". Changing the MaxNodes of a partition has no effect
1420 upon jobs that have already begun execution.
1421
1422
1423 MaxTime=<time>
1424 The maximum run time for jobs. Output format is
1425 [days-]hours:minutes:seconds or "UNLIMITED". Input format (for
1426 update command) is minutes, minutes:seconds, hours:minutes:sec‐
1427 onds, days-hours, days-hours:minutes or days-hours:minutes:sec‐
1428 onds. Time resolution is one minute and second values are
1429 rounded up to the next minute. Changing the MaxTime of a parti‐
1430 tion has no effect upon jobs that have already begun execution.
1431
1432
1433 MinNodes=<count>
1434 Set the minimum number of nodes which will be allocated to any
1435 single job in the partition. Changing the MinNodes of a parti‐
1436 tion has no effect upon jobs that have already begun execution.
1437
1438
1439 Nodes=<name>
1440 Identify the node(s) to be associated with this partition. Mul‐
1441 tiple node names may be specified using simple node range
1442 expressions (e.g. "lx[10-20]"). Note that jobs may only be
1443 associated with one partition at any time. Specify a blank data
1444 value to remove all nodes from a partition: "Nodes=". Changing
1445 the Nodes in a partition has no effect upon jobs that have
1446 already begun execution.
1447
1448
1449 OverTimeLimit=<count>
1450 Number of minutes by which a job can exceed its time limit
1451 before being canceled. The configured job time limit is treated
1452 as a soft limit. Adding OverTimeLimit to the soft limit pro‐
1453 vides a hard limit, at which point the job is canceled. This is
1454 particularly useful for backfill scheduling, which bases upon
1455 each job's soft time limit. A partition-specific OverTimeLimit
1456 will override any global OverTimeLimit value. If not specified,
1457 the global OverTimeLimit value will take precedence. May not
1458 exceed exceed 65533 minutes. An input value of "UNLIMITED" will
1459 clear any previously configured partition-specific OverTimeLimit
1460 value.
1461
1462
1463 OverSubscribe=<yes|no|exclusive|force>[:<job_count>]
1464 Specify if compute resources (i.e. individual CPUs) in this par‐
1465 tition can be shared by multiple jobs. Possible values are
1466 "YES", "NO", "EXCLUSIVE" and "FORCE". An optional job count
1467 specifies how many jobs can be allocated to use each resource.
1468
1469
1470 PartitionName=<name>
1471 Identify the partition to be updated. This specification is
1472 required.
1473
1474
1475 PreemptMode=<mode>
1476 Reset the mechanism used to preempt jobs in this partition if
1477 PreemptType is configured to preempt/partition_prio. The default
1478 preemption mechanism is specified by the cluster-wide Preempt‐
1479 Mode configuration parameter. Possible values are "OFF", "CAN‐
1480 CEL", "CHECKPOINT", "REQUEUE" and "SUSPEND".
1481
1482
1483 Priority=<count>
1484 Jobs submitted to a higher priority partition will be dispatched
1485 before pending jobs in lower priority partitions and if possible
1486 they will preempt running jobs from lower priority partitions.
1487 Note that a partition's priority takes precedence over a job's
1488 priority. The value may not exceed 65533.
1489
1490
1491 PriorityJobFactor=<count>
1492 Partition factor used by priority/multifactor plugin in calcu‐
1493 lating job priority. The value may not exceed 65533. Also see
1494 PriorityTier.
1495
1496
1497 PriorityTier=<count>
1498 Jobs submitted to a partition with a higher priority tier value
1499 will be dispatched before pending jobs in partition with lower
1500 priority tier value and, if possible, they will preempt
1501 running jobs from partitions with lower priority tier values.
1502 Note that a partition's priority tier takes precedence over a
1503 job's priority. The value may not exceed 65533. Also see Pri‐
1504 orityJobFactor.
1505
1506
1507 QOS=<QOSname|blank to remove>
1508 Set the partition QOS with a QOS name or to remove the Partition
1509 QOS leave the option blank.
1510
1511
1512 RootOnly=<yes|no>
1513 Specify if only allocation requests initiated by user root will
1514 be satisfied. This can be used to restrict control of the par‐
1515 tition to some meta-scheduler. Possible values are "YES" and
1516 "NO".
1517
1518
1519 ReqResv=<yes|no>
1520 Specify if only allocation requests designating a reservation
1521 will be satisfied. This is used to restrict partition usage to
1522 be allowed only within a reservation. Possible values are "YES"
1523 and "NO".
1524
1525
1526 Shared=<yes|no|exclusive|force>[:<job_count>]
1527 Renamed to OverSubscribe, see option descriptions above.
1528
1529
1530 State=<up|down|drain|inactive>
1531 Specify if jobs can be allocated nodes or queued in this parti‐
1532 tion. Possible values are "UP", "DOWN", "DRAIN" and "INACTIVE".
1533
1534 UP Designates that new jobs may queued on the partition,
1535 and that jobs may be allocated nodes and run from the
1536 partition.
1537
1538 DOWN Designates that new jobs may be queued on the parti‐
1539 tion, but queued jobs may not be allocated nodes and
1540 run from the partition. Jobs already running on the
1541 partition continue to run. The jobs must be explicitly
1542 canceled to force their termination.
1543
1544 DRAIN Designates that no new jobs may be queued on the par‐
1545 tition (job submission requests will be denied with an
1546 error message), but jobs already queued on the parti‐
1547 tion may be allocated nodes and run. See also the
1548 "Alternate" partition specification.
1549
1550 INACTIVE Designates that no new jobs may be queued on the par‐
1551 tition, and jobs already queued may not be allocated
1552 nodes and run. See also the "Alternate" partition
1553 specification.
1554
1555
1556 TRESBillingWeights=<TRES Billing Weights>
1557 TRESBillingWeights is used to define the billing weights of each
1558 TRES type that will be used in calculating the usage of a job.
1559 The calculated usage is used when calculating fairshare and when
1560 enforcing the TRES billing limit on jobs. Updates affect new
1561 jobs and not existing jobs. See the slurm.conf man page for
1562 more information.
1563
1564
1565
1566 SPECIFICATIONS FOR UPDATE COMMAND, POWERCAP
1567
1568
1569 PowerCap=<count>
1570 Set the amount of watts the cluster is limited to. Specify a
1571 number, "INFINITE" to enable the power capping logic without
1572 power restriction or "0" to disable the power capping logic.
1573 Update slurm.conf with any changes meant to be persistent across
1574 normal restarts of slurmctld or the execution of scontrol recon‐
1575 fig.
1576
1577
1578 SPECIFICATIONS FOR CREATE, UPDATE, AND DELETE COMMANDS, RESERVATIONS
1579
1580
1581
1582 Reservation=<name>
1583 Identify the name of the reservation to be created,
1584 updated, or deleted. This parameter is required for
1585 update and is the only parameter for delete. For create,
1586 if you do not want to give a reservation name, use "scon‐
1587 trol create res ..." and a name will be created automati‐
1588 cally.
1589
1590
1591 Accounts=<account list>
1592 List of accounts permitted to use the reserved nodes, for
1593 example "Accounts=physcode1,physcode2". A user in any of
1594 the accounts may use the reserved nodes. A new reserva‐
1595 tion must specify Users and/or Accounts. If both Users
1596 and Accounts are specified, a job must match both in
1597 order to use the reservation. Accounts can also be
1598 denied access to reservations by preceding all of the
1599 account names with '-'. Alternately precede the equal
1600 sign with '-'. For example,
1601 "Accounts=-physcode1,-physcode2" or
1602 "Accounts-=physcode1,physcode2" will permit any account
1603 except physcode1 and physcode2 to use the reservation.
1604 You can add or remove individual accounts from an exist‐
1605 ing reservation by using the update command and adding a
1606 '+' or '-' sign before the '=' sign. If accounts are
1607 denied access to a reservation (account name preceded by
1608 a '-'), then all other accounts are implicitly allowed to
1609 use the reservation and it is not possible to also
1610 explicitly specify allowed accounts.
1611
1612
1613 BurstBuffer=<buffer_spec>[,<buffer_spec>,...]
1614 Specification of burst buffer resources which are to be
1615 reserved. "buffer_spec" consists of four elements:
1616 [plugin:][type:]#[units] "plugin" is the burst buffer
1617 plugin name, currently either "cray" or "generic". If no
1618 plugin is specified, the reservation applies to all con‐
1619 figured burst buffer plugins. "type" specifies a Cray
1620 generic burst buffer resource, for example "nodes". if
1621 "type" is not specified, the number is a measure of stor‐
1622 age space. The "units" may be "N" (nodes), "K|KiB",
1623 "M|MiB", "G|GiB", "T|TiB", "P|PiB" (for powers of 1024)
1624 and "KB", "MB", "GB", "TB", "PB" (for powers of 1000).
1625 The default units are bytes for reservations of storage
1626 space. For example "BurstBuffer=cray:2TB" (reserve 2TB
1627 of storage plus 3 nodes from the Cray plugin) or "Burst‐
1628 Buffer=100GB" (reserve 100 GB of storage from all config‐
1629 ured burst buffer plugins). Jobs using this reservation
1630 are not restricted to these burst buffer resources, but
1631 may use these reserved resources plus any which are gen‐
1632 erally available. NOTE: Usually Slurm interprets KB, MB,
1633 GB, TB, PB, TB units as powers of 1024, but for Burst
1634 Buffers size specifications Slurm supports both IEC/SI
1635 formats. This is because the CRAY API for managing
1636 DataWarps supports both formats.
1637
1638
1639 CoreCnt=<num>
1640 This option is only supported when Select‐
1641 Type=select/cons_res. Identify number of cores to be
1642 reserved. If NodeCnt is used without the FIRST_CORES
1643 flag, this is the total number of cores to reserve where
1644 cores per node is CoreCnt/NodeCnt. If a nodelist is
1645 used, or if NodeCnt is used with the FIRST_CORES flag,
1646 this should be an array of core numbers by node:
1647 Nodes=node[1-5] CoreCnt=2,2,3,3,4 or flags=FIRST_CORES
1648 NodeCnt=5 CoreCnt=1,2,1,3,2.
1649
1650
1651 Licenses=<license>
1652 Specification of licenses (or other resources available
1653 on all nodes of the cluster) which are to be reserved.
1654 License names can be followed by a colon and count (the
1655 default count is one). Multiple license names should be
1656 comma separated (e.g. "Licenses=foo:4,bar"). A new
1657 reservation must specify one or more resource to be
1658 included: NodeCnt, Nodes and/or Licenses. If a reserva‐
1659 tion includes Licenses, but no NodeCnt or Nodes, then the
1660 option Flags=LICENSE_ONLY must also be specified. Jobs
1661 using this reservation are not restricted to these
1662 licenses, but may use these reserved licenses plus any
1663 which are generally available.
1664
1665
1666 NodeCnt=<num>[,num,...]
1667 Identify number of nodes to be reserved. The number can
1668 include a suffix of "k" or "K", in which case the number
1669 specified is multiplied by 1024. A new reservation must
1670 specify one or more resource to be included: NodeCnt,
1671 Nodes and/or Licenses.
1672
1673
1674 Nodes=<name>
1675 Identify the node(s) to be reserved. Multiple node names
1676 may be specified using simple node range expressions
1677 (e.g. "Nodes=lx[10-20]"). Specify a blank data value to
1678 remove all nodes from a reservation: "Nodes=". A new
1679 reservation must specify one or more resource to be
1680 included: NodeCnt, Nodes and/or Licenses. A specification
1681 of "ALL" will reserve all nodes. Set Flags=PART_NODES and
1682 PartitionName= in order for changes in the nodes associ‐
1683 ated with a partition to also be reflected in the nodes
1684 associated with a reservation.
1685
1686
1687 StartTime=<time_spec>
1688 The start time for the reservation. A new reservation
1689 must specify a start time. It accepts times of the form
1690 HH:MM:SS for a specific time of day (seconds are
1691 optional). (If that time is already past, the next day
1692 is assumed.) You may also specify midnight, noon, fika
1693 (3 PM) or teatime (4 PM) and you can have a time-of-day
1694 suffixed with AM or PM for running in the morning or the
1695 evening. You can also say what day the job will be run,
1696 by specifying a date of the form MMDDYY or MM/DD/YY or
1697 MM.DD.YY, or a date and time as YYYY-MM-DD[THH:MM[:SS]].
1698 You can also give times like now + count time-units,
1699 where the time-units can be minutes, hours, days, or
1700 weeks and you can tell Slurm to run the job today with
1701 the keyword today and to run the job tomorrow with the
1702 keyword tomorrow. You cannot update the StartTime of a
1703 reservation in ACTIVE state.
1704
1705
1706 EndTime=<time_spec>
1707 The end time for the reservation. A new reservation must
1708 specify an end time or a duration. Valid formats are the
1709 same as for StartTime.
1710
1711
1712 Duration=<time>
1713 The length of a reservation. A new reservation must
1714 specify an end time or a duration. Valid formats are
1715 minutes, minutes:seconds, hours:minutes:seconds,
1716 days-hours, days-hours:minutes, days-hours:minutes:sec‐
1717 onds, or UNLIMITED. Time resolution is one minute and
1718 second values are rounded up to the next minute. Output
1719 format is always [days-]hours:minutes:seconds.
1720
1721
1722 PartitionName=<name>
1723 Identify the partition to be reserved.
1724
1725
1726 Flags=<flags>
1727 Flags associated with the reservation. You can add or
1728 remove individual flags from an existing reservation by
1729 adding a '+' or '-' sign before the '=' sign. For exam‐
1730 ple: Flags-=DAILY (NOTE: this shortcut is not supported
1731 for all flags). Currently supported flags include:
1732
1733 ANY_NODES This is a reservation for burst buffers
1734 and/or licenses only and not compute nodes.
1735 If this flag is set, a job using this
1736 reservation may use the associated burst
1737 buffers and/or licenses plus any compute
1738 nodes. If this flag is not set, a job
1739 using this reservation may use only the
1740 nodes and licenses associated with the
1741 reservation.
1742
1743 DAILY Repeat the reservation at the same time
1744 every day.
1745
1746 FLEX Permit jobs requesting the reservation to
1747 begin prior to the reservation's start
1748 time, end after the reservation's end time,
1749 and use any resources inside and/or outside
1750 of the reservation regardless of any con‐
1751 straints possibly set in the reservation. A
1752 typical use case is to prevent jobs not
1753 explicitly requesting the reservation from
1754 using those reserved resources rather than
1755 forcing jobs requesting the reservation to
1756 use those resources in the time frame
1757 reserved. Another use case could be to
1758 always have a particular number of nodes
1759 with a specific feature reserved for a spe‐
1760 cific account so users in this account may
1761 use this nodes plus possibly other nodes
1762 without this feature.
1763
1764 FIRST_CORES Use the lowest numbered cores on a node
1765 only.
1766
1767 IGNORE_JOBS Ignore currently running jobs when creating
1768 the reservation. This can be especially
1769 useful when reserving all nodes in the sys‐
1770 tem for maintenance.
1771
1772 LICENSE_ONLY See ANY_NODES.
1773
1774 MAINT Maintenance mode, receives special account‐
1775 ing treatment. This partition is permitted
1776 to use resources that are already in
1777 another reservation.
1778
1779 NO_HOLD_JOBS_AFTER
1780 By default, when a reservation ends the
1781 reservation request will be removed from
1782 any pending jobs submitted to the reserva‐
1783 tion and will be put into a held state.
1784 Use this flag to let jobs run outside of
1785 the reservation after the reservation is
1786 gone.
1787
1788 OVERLAP This reservation can be allocated resources
1789 that are already in another reservation.
1790
1791 PART_NODES This flag can be used to reserve all nodes
1792 within the specified partition. Partition‐
1793 Name and Nodes=ALL must be specified or
1794 this option is ignored.
1795
1796 PURGE_COMP Purge the reservation once the last associ‐
1797 ated job has completed. Once the reserva‐
1798 tion has been created, it must be populated
1799 within 5 minutes of its start time or it
1800 will be purged before any jobs have been
1801 run.
1802
1803 REPLACE Nodes which are DOWN, DRAINED, or allocated
1804 to jobs are automatically replenished using
1805 idle resources. This option can be used to
1806 maintain a constant number of idle
1807 resources available for pending jobs (sub‐
1808 ject to availability of idle resources).
1809 This should be used with the NodeCnt reser‐
1810 vation option; do not identify specific
1811 nodes to be included in the reservation.
1812
1813 REPLACE_DOWN Nodes which are DOWN or DRAINED are auto‐
1814 matically replenished using idle resources.
1815 This option can be used to maintain a con‐
1816 stant sized pool of resources available for
1817 pending jobs (subject to availability of
1818 idle resources). This should be used with
1819 the NodeCnt reservation option; do not
1820 identify specific nodes to be included in
1821 the reservation.
1822
1823 SPEC_NODES Reservation is for specific nodes (output
1824 only)
1825
1826 STATIC_ALLOC Make it so after the nodes are selected for
1827 a reservation they don't change. Without
1828 this option when nodes are selected for a
1829 reservation and one goes down the reserva‐
1830 tion will select a new node to fill the
1831 spot.
1832
1833 TIME_FLOAT The reservation start time is relative to
1834 the current time and moves forward through
1835 time (e.g. a StartTime=now+10minutes will
1836 always be 10 minutes in the future).
1837
1838 WEEKDAY Repeat the reservation at the same time on
1839 every weekday (Monday, Tuesday, Wednesday,
1840 Thursday and Friday).
1841
1842 WEEKEND Repeat the reservation at the same time on
1843 every weekend day (Saturday and Sunday).
1844
1845 WEEKLY Repeat the reservation at the same time
1846 every week.
1847
1848
1849 Features=<features>
1850 Set the reservation's required node features. Multiple
1851 values may be "&" separated if all features are required
1852 (AND operation) or separated by "|" if any of the speci‐
1853 fied features are required (OR operation). Parenthesis
1854 are also supported for features to be ANDed together with
1855 counts of nodes having the specified features. For exam‐
1856 ple "Features=[(knl&a2a&flat)*4&haswell*2]" indicates the
1857 advanced reservation should include 4 nodes with ALL of
1858 the features "knl", "a2a", and "flat" plus 2 nodes with
1859 the feature "haswell".
1860
1861 Value may be cleared with blank data value, "Features=".
1862
1863
1864 Users=<user list>
1865 List of users permitted to use the reserved nodes, for
1866 example "User=jones1,smith2". A new reservation must
1867 specify Users and/or Accounts. If both Users and
1868 Accounts are specified, a job must match both in order to
1869 use the reservation. Users can also be denied access to
1870 reservations by preceding all of the user names with '-'.
1871 Alternately precede the equal sign with '-'. For exam‐
1872 ple, "User=-jones1,-smith2" or "User-=jones1,smith2" will
1873 permit any user except jones1 and smith2 to use the
1874 reservation. You can add or remove individual users from
1875 an existing reservation by using the update command and
1876 adding a '+' or '-' sign before the '=' sign. If users
1877 are denied access to a reservation (user name preceded by
1878 a '-'), then all other users are implicitly allowed to
1879 use the reservation and it is not possible to also
1880 explicitly specify allowed users.
1881
1882
1883 TRES=<tres_spec>
1884 Comma-separated list of TRES required for the reserva‐
1885 tion. Current supported TRES types with reservations are:
1886 CPU, Node, License and BB. CPU and Node follow the same
1887 format as CoreCnt and NodeCnt parameters respectively.
1888 License names can be followed by an equal '=' and a
1889 count:
1890
1891 License/<name1>=<count1>[,License/<name2>=<count2>,...]
1892
1893 BurstBuffer can be specified in a similar way as Burst‐
1894 Buffer parameter. The only difference is that colon sym‐
1895 bol ':' should be replaced by an equal '=' in order to
1896 follow the TRES format.
1897
1898 Some examples of TRES valid specifications:
1899
1900 TRES=cpu=5,bb/cray=4,license/iop1=1,license/iop2=3
1901
1902 TRES=node=5k,license/iop1=2
1903
1904 As specified in CoreCnt, if a nodelist is specified, cpu
1905 can be an array of core numbers by node: nodes=com‐
1906 pute[1-3] TRES=cpu=2,2,1,bb/cray=4,license/iop1=2
1907
1908 Please note that CPU, Node, License and BB can override
1909 CoreCnt, NodeCnt, Licenses and BurstBuffer parameters
1910 respectively. Also CPU represents CoreCnt, in a reserva‐
1911 tion and will be adjusted if you have threads per core on
1912 your nodes.
1913
1914
1915 SPECIFICATIONS FOR SHOW COMMAND, LAYOUTS
1916
1917 Without options, lists all configured layouts. With a layout
1918 specified, shows entities with following options:
1919
1920
1921 Key=<value>
1922 Keys/Values to update for the entities. The format must
1923 respect the layout.d configuration files. Key=Type cannot
1924 be updated. One Key/Value is required, several can be
1925 set.
1926
1927 Entity=<value>
1928 Entities to show, default is not used. Can be set to "*".
1929
1930 Type=<value>
1931 Type of entities to show, default is not used.
1932
1933 nolayout
1934 If not used, only entities with defining the tree are
1935 shown. With the option, only leaves are shown.
1936
1937
1939 Some scontrol options may be set via environment variables.
1940 These environment variables, along with their corresponding
1941 options, are listed below. (Note: Commandline options will
1942 always override these settings.)
1943
1944 SCONTROL_ALL -a, --all
1945
1946 SCONTROL_FEDERATION --federation
1947
1948 SCONTROL_FUTURE -F, --future
1949
1950 SCONTROL_LOCAL --local
1951
1952 SCONTROL_SIBLING --sibling
1953
1954 SLURM_BITSTR_LEN Specifies the string length to be used for
1955 holding a job array's task ID expression.
1956 The default value is 64 bytes. A value of 0
1957 will print the full expression with any
1958 length required. Larger values may
1959 adversely impact the application perfor‐
1960 mance.
1961
1962 SLURM_CLUSTERS Same as --clusters
1963
1964 SLURM_CONF The location of the Slurm configuration
1965 file.
1966
1967 SLURM_TIME_FORMAT Specify the format used to report time
1968 stamps. A value of standard, the default
1969 value, generates output in the form
1970 "year-month-dateThour:minute:second". A
1971 value of relative returns only
1972 "hour:minute:second" if the current day.
1973 For other dates in the current year it
1974 prints the "hour:minute" preceded by
1975 "Tomorr" (tomorrow), "Ystday" (yesterday),
1976 the name of the day for the coming week
1977 (e.g. "Mon", "Tue", etc.), otherwise the
1978 date (e.g. "25 Apr"). For other years it
1979 returns a date month and year without a time
1980 (e.g. "6 Jun 2012"). All of the time stamps
1981 use a 24 hour format.
1982
1983 A valid strftime() format can also be speci‐
1984 fied. For example, a value of "%a %T" will
1985 report the day of the week and a time stamp
1986 (e.g. "Mon 12:34:56").
1987
1988
1989 SLURM_TOPO_LEN Specify the maximum size of the line when
1990 printing Topology. If not set, the default
1991 value "512" will be used.
1992
1993
1995 When using the Slurm db, users who have AdminLevel's defined
1996 (Operator or Admin) and users who are account coordinators are
1997 given the authority to view and modify jobs, reservations,
1998 nodes, etc., as defined in the following table - regardless of
1999 whether a PrivateData restriction has been defined in the
2000 slurm.conf file.
2001
2002 scontrol show job(s): Admin, Operator, Coordinator
2003 scontrol update job: Admin, Operator, Coordinator
2004 scontrol requeue: Admin, Operator, Coordinator
2005 scontrol show step(s): Admin, Operator, Coordinator
2006 scontrol update step: Admin, Operator, Coordinator
2007
2008 scontrol show node: Admin, Operator
2009 scontrol update node: Admin
2010
2011 scontrol create partition: Admin
2012 scontrol show partition: Admin, Operator
2013 scontrol update partition: Admin
2014 scontrol delete partition: Admin
2015
2016 scontrol create reservation: Admin, Operator
2017 scontrol show reservation: Admin, Operator
2018 scontrol update reservation: Admin, Operator
2019 scontrol delete reservation: Admin, Operator
2020
2021 scontrol reconfig: Admin
2022 scontrol shutdown: Admin
2023 scontrol takeover: Admin
2024
2025
2027 # scontrol
2028 scontrol: show part debug
2029 PartitionName=debug
2030 AllocNodes=ALL AllowGroups=ALL Default=YES
2031 DefaultTime=NONE DisableRootJobs=NO Hidden=NO
2032 MaxNodes=UNLIMITED MaxTime=UNLIMITED MinNodes=1
2033 Nodes=snowflake[0-48]
2034 Priority=1 RootOnly=NO OverSubscribe=YES:4
2035 State=UP TotalCPUs=694 TotalNodes=49
2036 scontrol: update PartitionName=debug MaxTime=60:00 MaxNodes=4
2037 scontrol: show job 71701
2038 JobId=71701 Name=hostname
2039 UserId=da(1000) GroupId=da(1000)
2040 Priority=66264 Account=none QOS=normal WCKey=*123
2041 JobState=COMPLETED Reason=None Dependency=(null)
2042 TimeLimit=UNLIMITED Requeue=1 Restarts=0 BatchFlag=0 Exit‐
2043 Code=0:0
2044 SubmitTime=2010-01-05T10:58:40 Eligible‐
2045 Time=2010-01-05T10:58:40
2046 StartTime=2010-01-05T10:58:40 EndTime=2010-01-05T10:58:40
2047 SuspendTime=None SecsPreSuspend=0
2048 Partition=debug AllocNode:Sid=snowflake:4702
2049 ReqNodeList=(null) ExcNodeList=(null)
2050 NodeList=snowflake0
2051 NumNodes=1 NumCPUs=10 CPUs/Task=2 ReqS:C:T=1:1:1
2052 MinCPUsNode=2 MinMemoryNode=0 MinTmpDiskNode=0
2053 Features=(null) Reservation=(null)
2054 OverSubscribe=OK Contiguous=0 Licenses=(null) Network=(null)
2055 scontrol: update JobId=71701 TimeLimit=30:00 Priority=500
2056 scontrol: show hostnames tux[1-3]
2057 tux1
2058 tux2
2059 tux3
2060 scontrol: create res StartTime=2009-04-01T08:00:00 Dura‐
2061 tion=5:00:00 Users=dbremer NodeCnt=10
2062 Reservation created: dbremer_1
2063 scontrol: update Reservation=dbremer_1 Flags=Maint NodeCnt=20
2064 scontrol: delete Reservation=dbremer_1
2065 scontrol: quit
2066
2067
2069 Copyright (C) 2002-2007 The Regents of the University of Cali‐
2070 fornia. Produced at Lawrence Livermore National Laboratory (cf,
2071 DISCLAIMER).
2072 Copyright (C) 2008-2010 Lawrence Livermore National Security.
2073 Copyright (C) 2010-2018 SchedMD LLC.
2074
2075 This file is part of Slurm, a resource management program. For
2076 details, see <https://slurm.schedmd.com/>.
2077
2078 Slurm is free software; you can redistribute it and/or modify it
2079 under the terms of the GNU General Public License as published
2080 by the Free Software Foundation; either version 2 of the
2081 License, or (at your option) any later version.
2082
2083 Slurm is distributed in the hope that it will be useful, but
2084 WITHOUT ANY WARRANTY; without even the implied warranty of MER‐
2085 CHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
2086 General Public License for more details.
2087
2089 /etc/slurm.conf
2090
2092 scancel(1), sinfo(1), squeue(1), slurm_checkpoint [22m(3),
2093 slurm_create_partition [22m(3), slurm_delete_partition [22m(3),
2094 slurm_load_ctl_conf [22m(3), slurm_load_jobs [22m(3), slurm_load_node
2095 (3), slurm_load_partitions [22m(3), slurm_reconfigure [22m(3),
2096 slurm_requeue [22m(3), slurm_resume [22m(3), slurm_shutdown [22m(3),
2097 slurm_suspend (3), slurm_takeover [22m(3), slurm_update_job [22m(3),
2098 slurm_update_node [22m(3), slurm_update_partition [22m(3),
2099 slurm.conf(5), slurmctld(8)
2100
2101
2102
2103February 2019 Slurm Commands scontrol(1)