SRUN(1) slurm components SRUN(1) NAME srun - run parallel jobs SYNOPSIS srun [OPTIONS...] executable [args...] DESCRIPTION Run a parallel job on cluster managed by SLURM. If necessary, srun will first create a resource allocation in which to run the parallel job. OPTIONS --acctg-freq=seconds Define the job accounting sampling interval. This can be used to override the JobAcctGatherFrequency parameter in SLURM's con- figuration file, slurm.conf. A value of zero disables real the periodic job sampling and provides accounting information only on job termination (reducing SLURM interference with the job). -B --extra-node-info=sockets[:cores[:threads]] Request a specific allocation of resources with details as to the number and type of computational resources within a cluster: number of sockets (or physical processors) per node, cores per socket, and threads per core. The total amount of resources being requested is the product of all of the terms. As with --nodes, each value can be a single number or a range (e.g. min-max). An asterisk (*) can be used as a placeholder indicat- ing that all available resources of that type are to be uti- lized. As with nodes, the individual levels can also be speci- fied in separate options if desired: --sockets-per-node=sockets --cores-per-socket=cores --threads-per-core=threads When the task/affinity plugin is enabled, specifying an alloca- tion in this manner also instructs SLURM to use a CPU affinity mask to guarantee the request is filled as specified. NOTE: Support for these options are configuration dependent. The task/affinity plugin must be configured. In addition either select/linear or select/cons_res plugin must be configured. If select/cons_res is configured, it must have a parameter of CR_Core, CR_Core_Memory, CR_Socket, or CR_Socket_Memory. --begin=time Defer initiation of this job until the specified 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, or teatime (4pm) 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. You can also give times like now + count time-units, where the time-units can be seconds (default), 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. The value may be changed after job submission using the scontrol command. For example: --begin=16:00 --begin=now+1hour --begin=now+60 (seconds by default) --begin=02/22/08-12:34:67 --checkpoint=time Specifies the interval between creating checkpoints of the job step. By default, the job step will no checkpoints created. Acceptable time formats include "minutes", "minutes:seconds", "hours:minutes:seconds", "days-hours", "days-hours:minutes" and "days-hours:minutes:seconds". --checkpoint-path=directory Specifies the directory into which the job step's checkpoint should be written (used by the checkpoint/xlch plugin only). -C, --constraint[=] Specify a list of constraints. The constraints are features that have been assigned to the nodes by the slurm administrator. The list of constraints may include multiple features separated by ampersand (AND) and/or vertical bar (OR) operators. For example: --constraint="opteron&video" or --con- straint="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 that case that 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 brack- ets. For example: "--constraint="[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 "srun --nodes=16 --constraint=graph- ics*4 ..." indicates that the job requires 16 nodes at that at least four of those nodes must have the feature "graphics." Constraints with node counts may only be combined with AND oper- ators. If no nodes have the requested features, then the job will be rejected by the slurm job manager. --cpu_bind=[{quiet,verbose},]type Bind tasks to CPUs. Used only when the task/affinity plugin is enabled. NOTE: To have SLURM always report on the selected CPU binding for all commands executed in a shell, you can enable verbose mode by setting the SLURM_CPU_BIND environment variable value to "verbose". Note that the SLURM_CPU_BIND environment variable will propagate into the tasks' environment in order to perform binding in batch submissions. If you do not wish to propagate --cpu_bind to suc- cessive srun commands, simply clear the variable in the task's script before executing srun: unsetenv SLURM_CPU_BIND In addition, to SLURM_CPU_BIND, the following informational environment variables are also set when SLURM_CPU_BIND is in use: SLURM_CPU_BIND_VERBOSE SLURM_CPU_BIND_TYPE SLURM_CPU_BIND_LIST See the ENVIRONMENT VARIABLE section for a more detailed description of the individual SLURM_CPU_BIND* variables. When using --cpus-per-task to run multithreaded tasks, be aware that CPU binding is inherited from the parent of the process. This means that the multithreaded task should either specify or clear the CPU binding itself to avoid having all threads of the multithreaded task use the same mask/CPU as the parent. Alter- natively, fat masks (masks which specify more than one allowed CPU) could be used for the tasks in order to provide multiple CPUs for the multithreaded tasks. Supported options include: q[uiet] quietly bind before task runs (default) v[erbose] verbosely report binding before task runs no[ne] don't bind tasks to CPUs (default) rank bind by task rank map_cpu: bind by mapping CPU IDs to tasks as specified where is ,,.... CPU IDs are interpreted as decimal values unless they are preceded with '0x' in which case they interpreted as hexadecimal values. mask_cpu: bind by setting CPU masks on tasks as specified where is ,,.... CPU masks are always interpreted as hexadecimal values but can be pre- ceded with an optional '0x'. sockets auto-generated masks bind to sockets cores auto-generated masks bind to cores threads auto-generated masks bind to threads help show this help message -c, --cpus-per-task=ncpus Request that ncpus be allocated per process. This may be useful if the job is multithreaded and requires more than one cpu per task for optimal performance. The default is one cpu per pro- cess. If -c is specified without -n as many tasks will be allo- cated per node as possible while satisfying the -c restriction. --comment= An arbitrary comment. --contiguous Demand a contiguous range of nodes. The default is "yes". Spec- ify --contiguous=no if a contiguous range of nodes is not a con- straint. --core=type Adjust corefile format for parallel job. If possible, srun will set up the environment for the job such that a corefile format other than full core dumps is enabled. If run with type = "list", srun will print a list of supported corefile format types to stdout and exit. --ctrl-comm-ifhn=addr Specify the address or hostname to be used for PMI communica- tions only (task communication and synchronization primitives for MPCIH2). Defaults to hostname (response from getnodename function). Use of this is required if a DNS lookup can not be performed on the hostname or if that address is blocked from the compute nodes. -D, --chdir=path have the remote processes do a chdir to path before beginning execution. The default is to chdir to the current working direc- tory of the srun process. -d, --slurmd-debug=level Specify a debug level for slurmd(8). level may be an integer value between 0 [quiet, only errors are displayed] and 4 [ver- bose operation]. The slurmd debug information is copied onto the stderr of the job. By default only errors are displayed. -e, --error=mode Specify how stderr is to be redirected. By default in interac- tive mode, srun redirects stderr to the same file as stdout, if one is specified. The --error option is provided to allow stdout and stderr to be redirected to different locations. See IO Redirection below for more options. If the specified file already exists, it will be overwritten. --epilog=executable srun will run executable just after the job step completes. The command line arguments for executable will be the command and arguments of the job step. If executable is "none", then no epilog will be run. This parameter overrides the SrunEpilog parameter in slurm.conf. --exclusive When used to initiate a job step within an existing resource allocation, proceed only when processors can be dedicated to the job step without sharing with other job steps. This can be used to initiate many job steps simultaneously within an existing job allocation and have SLURM perform resource management for the job. In this mode, use with the --ntasks option and NOT the --nodes, --relative, --relative=arbitrary options (which provide user control over task layout). See EXAMPLE below. When used to initiate a job, dedicate whole nodes to the job rather than individual processors even if consumable resources are enabled (e.g. SelectType=select/cons_res). --gid=group If srun is run as root, and the --gid option is used, submit the job with group's group access permissions. group may be the group name or the numerical group ID. --help Display verbose help message and exit. --hint=type Bind tasks according to application hints compute_bound Select settings for compute bound applications: use all cores in each physical CPU memory_bound Select settings for memory bound applications: use only one core in each physical CPU [no]multithread [don't] use extra threads with in-core multi-threading which can benefit communication intensive applications help show this help message -I, --immediate exit if resources are not immediately available. By default, --immediate is off, and srun will block until resources become available. -i, --input=mode Specify how stdin is to redirected. By default, srun redirects stdin from the terminal all tasks. See IO Redirection below for more options. For OS X, the poll() function does not support stdin, so input from a terminal is not possible. -J, --job-name=jobname Specify a name for the job. The specified name will appear along with the job id number when querying running jobs on the system. The default is the supplied executable program's name. --jobid=id Initiate a job step under an already allocated job with job id id. Using this option will cause srun to behave exactly as if the SLURM_JOBID environment variable was set. -K, --kill-on-bad-exit Terminate a job if any task exits with a non-zero exit code. -k, --no-kill Do not automatically terminate a job of one of the nodes it has been allocated fails. This option is only recognized on a job allocation, not for the submission of individual job steps. The job will assume all responsibilities for fault-tolerance. The active job step (MPI job) will almost certainly suffer a fatal error, but subsequent job steps may be run if this option is specified. The default action is to terminate job upon node failure. -l, --label prepend task number to lines of stdout/err. Normally, stdout and stderr from remote tasks is line-buffered directly to the stdout and stderr of srun. The --label option will prepend lines of output with the remote task id. -L, --licenses= Specification of licenses (or other resources available on all nodes of the cluster) which must be allocated to this job. License names can be followed by an asterisk and count (the default count is one). Multiple license names should be comma separated (e.g. "--licenses=foo*4,bar"). -m, --relative (block|cyclic|arbitrary|plane=) Specify an alternate distribution method for remote processes. block The block method of distribution will allocate processes in-order to the cpus on a node. If the number of pro- cesses exceeds the number of cpus on all of the nodes in the allocation then all nodes will be utilized. For exam- ple, consider an allocation of three nodes each with two cpus. A four-process block distribution request will dis- tribute those processes to the nodes with processes one and two on the first node, process three on the second node, and process four on the third node. Block distri- bution is the default behavior if the number of tasks exceeds the number of nodes requested. cyclic The cyclic method distributes processes in a round-robin fashion across the allocated nodes. That is, process one will be allocated to the first node, process two to the second, and so on. This is the default behavior if the number of tasks is no larger than the number of nodes requested. plane The tasks are distributed in blocks of a specified size. The options include a number representing the size of the task block. This is followed by an optional specifica- tion of the task distribution scheme within a block of tasks and between the blocks of tasks. For more details (including examples and diagrams), please see https://computing.llnl.gov/linux/slurm/mc_support.html and https://computing.llnl.gov/linux/slurm/dist_plane.html. arbitrary The arbitrary method of distribution will allocate pro- cesses in-order as listed in file designated by the envi- ronment variable SLURM_HOSTFILE. If this variable is listed it will over ride any other method specified. If not set the method will default to block. Inside the hostfile must contain at minimum the number of hosts requested. If requesting tasks (-n) your tasks will be laid out on the nodes in the order of the file. --mail-type=type Notify user by email when certain event types occur. Valid type values are BEGIN, END, FAIL, ALL (any state change). The user to be notified is indicated with --mail-user. --mail-user=user User to receive email notification of state changes as defined by --mail-type. The default value is the submitting user. --mem[=] Specify the real memory required per node in MegaBytes. Default value is DefMemPerNode and the maximum value is MaxMemPerNode. If configured, both of parameters can be seen using the scontrol show config command. This parameter would generally be used of whole nodes are allocated to jobs (SelectType=select/linear). Also see --mem-per-cpu. --mem and --mem-per-cpu are mutually exclusive. --mem-per-cpu[=] Mimimum memory required per allocated CPU in MegaBytes. Default value is DefMemPerCPU and the maximum value is MaxMemPerCPU. If configured, both of parameters can be seen using the scontrol show config command. This parameter would generally be used of individual processors are allocated to jobs (Select- Type=select/cons_res). Also see --mem. --mem and --mem-per-cpu are mutually exclusive. --mem_bind=[{quiet,verbose},]type Bind tasks to memory. Used only when the task/affinity plugin is enabled and the NUMA memory functions are available. Note that the resolution of CPU and memory binding may differ on some architectures. For example, CPU binding may be performed at the level of the cores within a processor while memory binding will be performed at the level of nodes, where the definition of "nodes" may differ from system to system. The use of any type other than "none" or "local" is not recommended. If you want greater control, try running a simple test code with the options "--cpu_bind=verbose,none --mem_bind=verbose,none" to determine the specific configuration. NOTE: To have SLURM always report on the selected memory binding for all commands executed in a shell, you can enable verbose mode by setting the SLURM_MEM_BIND environment variable value to "verbose". Note that the SLURM_MEM_BIND environment variable will propagate into the tasks' environment in order to perform binding in batch submissions. If you do not wish to propagate --mem_bind to suc- cessive srun commands, simply clear the variable in the task's script before executing srun: unsetenv SLURM_MEM_BIND In addition, to SLURM_MEM_BIND, the following informational environment variables are also set when SLURM_MEM_BIND is in use: SLURM_MEM_BIND_VERBOSE SLURM_MEM_BIND_TYPE SLURM_MEM_BIND_LIST See the ENVIRONMENT VARIABLES section for a more detailed description of the individual SLURM_MEM_BIND* variables. Supported options include: q[uiet] quietly bind before task runs (default) v[erbose] verbosely report binding before task runs no[ne] don't bind tasks to memory (default) rank bind by task rank (not recommended) local Use memory local to the processor in use map_mem: bind by mapping a node's memory to tasks as specified where is ,,.... CPU IDs are interpreted as decimal values unless they are pre- ceded with '0x' in which case they interpreted as hex- adecimal values (not recommended) mask_mem: bind by setting memory masks on tasks as specified where is ,,.... memory masks are always interpreted as hexadecimal values. Note that masks must be preceded with a '0x' if they don't begin with [0-9] so they are seen as numerical values by srun. help show this help message --mincpus=n Specify a minimum number of logical cpus/processors per node. --minsockets=n Specify a minimum number of sockets (physical processors) per node. --mincores=n Specify a minimum number of cores per socket. --minthreads=n Specify a minimum number of threads per core. --msg-timeout=seconds Modify the job launch message timeout. The default value is MessageTimeout in the SLURM configuration file slurm.conf. Changes to this are typically not recommended, but could be use- ful to diagnose problems. --mpi=mpi_type Identify the type of MPI to be used. May result in unique initi- ation procedures. list Lists available mpi types to choose from. lam Initiates one 'lamd' process per node and establishes necessary environment variables for LAM/MPI. mpich1_shmem Initiates one process per node and establishes necessary environment variables for mpich1 shared memory model. This also works for mvapich built for shared memory. mpichgm For use with Myrinet. mvapich For use with Infiniband. openmpi For use with OpenMPI. none No special MPI processing. This is the default and works with many other versions of MPI. --multi-prog Run a job with different programs and different arguments for each task. In this case, the executable program specified is actually a configuration file specifying the executable and arguments for each task. See MULTIPLE PROGRAM CONFIGURATION below for details on the configuration file contents. -N, --nodes=minnodes[-maxnodes] Request that a minimum of minnodes nodes be allocated to this job. The scheduler may decide to launch the job on more than minnodes nodes. A limit on the maximum node count may be speci- fied with maxnodes (e.g. "--nodes=2-4"). The minimum and maxi- mum node count may be the same to specify a specific number of nodes (e.g. "--nodes=2-2" will ask for two and ONLY two nodes). The partition's node limits supersede those of the job. If a job's node limits are outside of the range permitted for its associated partition, the job will be left in a PENDING state. This permits possible execution at a later time, when the parti- tion limit is changed. If a job node limit exceeds the number of nodes configured in the partition, the job will be rejected. Note that the environment variable SLURM_NNODES will be set to the count of nodes actually allocated to the job. See the ENVI- RONMENT VARIABLES section for more information. If -N is not specified, the default behavior is to allocate enough nodes to satisfy the requirements of the -n and -c options. The job will be allocated as many nodes as possible within the range speci- fied and without delaying the initiation of the job. -n, --ntasks=ntasks Specify the number of tasks to run. Request that srun allocate resources for ntasks tasks. The default is one task per socket or core (depending upon the value of the SelectTypeParameters parameter in slurm.conf), but note that the --cpus-per-task option will change this default. --network=type Specify the communication protocol to be used. This option is supported on AIX systems. Since POE is used to launch tasks, this option is not normally used or is specified using the SLURM_NETWORK environment variable. The interpretation of type is system dependent. For systems with an IBM Federation switch, the following comma-separated and case insensitive types are recognized: IP (the default is user-space), SN_ALL, SN_SINGLE, BULK_XFER and adapter names (e.g. SNI0 and SNI1). For more information, on IBM systems see poe documentation on the envi- ronment variables MP_EUIDEVICE and MP_USE_BULK_XFER. Note that only four jobs steps may be active at once on a node with the BULK_XFER option due to limitations in the Federation switch driver. --nice[=adjustment] Run the job with an adjusted scheduling priority within SLURM. With no adjustment value the scheduling priority is decreased by 100. The adjustment range is from -10000 (highest priority) to 10000 (lowest priority). Only privileged users can specify a negative adjustment. NOTE: This option is presently ignored if SchedulerType=sched/wiki or SchedulerType=sched/wiki2. --ntasks-per-core=ntasks Request that no more than ntasks be invoked on each core. Simi- lar to --ntasks-per-node except at the core level instead of the node level. Masks will automatically be generated to bind the tasks to specific core unless --cpu_bind=none is specified. NOTE: This option is not supported unless SelectType=CR_Core or SelectType=CR_Core_Memory is configured. --ntasks-per-socket=ntasks Request that no more than ntasks be invoked on each socket. Similar to --ntasks-per-node except at the socket level instead of the node level. Masks will automatically be generated to bind the tasks to specific sockets unless --cpu_bind=none is specified. NOTE: This option is not supported unless Select- Type=CR_Socket or SelectType=CR_Socket_Memory is configured. --ntasks-per-node=ntasks Request that no more than ntasks be invoked on each node. This is similar to using --cpus-per-task=ncpus but does not require knowledge of the actual number of cpus on each node. In some cases, it is more convenient to be able to request that no more than a specific number of ntasks be invoked on each node. Exam- ples of this include submitting a hybrid MPI/OpenMP app where only one MPI "task/rank" should be assigned to each node while allowing the OpenMP portion to utilize all of the parallelism present in the node, or submitting a single setup/cleanup/moni- toring job to each node of a pre-existing allocation as one step in a larger job script. -O, --overcommit overcommit resources. Normally, srun will not allocate more than one process per cpu. By specifying --overcommit you are explic- itly allowing more than one process per cpu. However no more than MAX_TASKS_PER_NODE tasks are permitted to execute per node. -o, --output=mode Specify the mode for stdout redirection. By default in interac- tive mode, srun collects stdout from all tasks and line buffers this output to the attached terminal. With --output stdout may be redirected to a file, to one file per task, or to /dev/null. See section IO Redirection below for the various forms of mode. If the specified file already exists, it will be overwritten. If --error is not also specified on the command line, both std- out and stderr will directed to the file specified by --output. --open-mode=append|truncate Open the output and error files using append or truncate mode as specified. The default value is specified by the system config- uration parameter JobFileAppend. -P, --dependency[=] Defer the start of this job until the specified dependencies have been satisfied completed. is of the form . Many jobs can share the same dependency and these jobs may even belong to dif- ferent users. The value may be changed after job submission using the scontrol command. 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 non- zero exit code). singleton This job can begin execution after any previously launched jobs sharing the same job name and user have terminated. -p, --partition=partition Request resources from partition "partition." Partitions are created by the slurm administrator, who also identify one of those partitions as the default. --prolog=executable srun will run executable just before launching the job step. The command line arguments for executable will be the command and arguments of the job step. If executable is "none", then no prolog will be run. This parameter overrides the SrunProlog parameter in slurm.conf. --propagate[=rlimits] Allows users to specify which of the modifiable (soft) resource limits to propagate to the compute nodes and apply to their jobs. If rlimits is not specified, then all resource limits will be propagated. The following rlimit names are supported by Slurm (although some options may not be supported on some sys- tems): ALL All limits listed below AS The maximum address space for a processes CORE The maximum size of core file CPU The maximum amount of CPU time DATA The maximum size of a process's data segment FSIZE The maximum size of files created MEMLOCK The maximum size that may be locked into memory NOFILE The maximum number of open files NPROC The maximum number of processes available RSS The maximum resident set size STACK The maximum stack size --pty Execute task zero in pseudo terminal. Implicitly sets --unbuffered. Implicitly sets --error and --output to /dev/null for all tasks except task zero. Not currently supported on AIX platforms. -Q, --quiet Quiet operation. Suppress informational messages. Errors will still be displayed. -q, --quit-on-interrupt Quit immediately on single SIGINT (Ctrl-C). Use of this option disables the status feature normally available when srun receives a single Ctrl-C and causes srun to instead immediately terminate the running job. -r, --relative=n Run a job step relative to node n of the current allocation. This option may be used to spread several job steps out among the nodes of the current job. If -r is used, the current job step will begin at node n of the allocated nodelist, where the first node is considered node 0. The -r option is not permitted along with -w or -x, and will be silently ignored when not run- ning within a prior allocation (i.e. when SLURM_JOBID is not set). The default for n is 0. If the value of --nodes exceeds the number of nodes identified with the --relative option, a warning message will be printed and the --relative option will take precedence. -s, --share The job can share nodes with other running jobs. This may result in faster job initiation and higher system utilization, but lower application performance. -T, --threads=nthreads Request that srun use nthreads to initiate and control the par- allel job. The default value is the smaller of 60 or the number of nodes allocated. This should only be used to set a low thread count for testing on very small memory computers. -t, --time=time Establish a time limit to terminate the job after the specified period of time. If the job's time limit exceeds the partition's time limit, the job will be left in a PENDING state. The default value is the partition's time limit. When the time limit is reached, the job's processes are sent SIGTERM followed by SIGKILL. The interval between signals is specified by the SLURM configuration parameter KillWait. A time limit of 0 minutes indicates that an infinite timelimit should be used. Acceptable time formats include "minutes", "minutes:seconds", "hours:min- utes:seconds", "days-hours", "days-hours:minutes" and "days-hours:minutes:seconds". --task-epilog=executable The slurmd daemon will run executable just after each task ter- minates. This will be before after any TaskEpilog parameter in slurm.conf is executed. This is meant to be a very short-lived program. If it fails to terminate within a few seconds, it will be killed along with any descendant processes. --task-prolog=executable The slurmd daemon will run executable just before launching each task. This will be executed after any TaskProlog parameter in slurm.conf is executed. Besides the normal environment vari- ables, this has SLURM_TASK_PID available to identify the process ID of the task being started. Standard output from this program of the form "export NAME=value" will be used to set environment variables for the task being spawned. --tmp=MB Specify a minimum amount of temporary disk space. -U, --account=account Change resource use by this job to specified account. The account is an arbitrary string. The may be changed after job submission using the scontrol command. -u, --unbuffered do not line buffer stdout from remote tasks. This option cannot be used with --label. --uid=user Attempt to submit and/or run a job as user instead of the invok- ing user id. The invoking user's credentials will be used to check access permissions for the target partition. User root may use this option to run jobs as a normal user in a RootOnly par- tition for example. If run as root, srun will drop its permis- sions to the uid specified after node allocation is successful. user may be the user name or numerical user ID. --usage Display brief help message and exit. -V, --version Display version information and exit. -v, --verbose Print detailed event logging. Multiple -v's will further increase the verbosity of logging. By default only errors will be displayed. -W, --wait=seconds Specify how long to wait after the first task terminates before terminating all remaining tasks. A value of 0 indicates an unlimited wait (a warning will be issued after 60 seconds). The default value is set by the WaitTime parameter in the slurm con- figuration file (see slurm.conf(5)). This option can be useful to insure that a job is terminated in a timely fashion in the event that one or more tasks terminate prematurely. -w, --nodelist=host1,host2,... or filename Request a specific list of hosts. The job will contain at least these hosts. The list may be specified as a comma-separated list of hosts, a range of hosts (host[1-5,7,...] for example), or a filename. The host list will be assumed to be a filename if it contains a "/" character. If you specify a max node count (-N1-2) if there are more than 2 hosts in the file only the first 2 nodes will be used in the request list. -X, --disable-status Disable the display of task status when srun receives a single SIGINT (Ctrl-C). Instead immediately forward the SIGINT to the running job. A second Ctrl-C in one second will forcibly termi- nate the job and srun will immediately exit. May also be set via the environment variable SLURM_DISABLE_STATUS. -x, --exclude=host1,host2,... or filename Request that a specific list of hosts not be included in the resources allocated to this job. The host list will be assumed to be a filename if it contains a "/"character. The following options support Blue Gene systems, but may be applicable to other systems as well. --blrts-image=path Path to blrts image for bluegene block. Default from blu- gene.conf if not set. --conn-type=type Require the partition connection type to be of a certain type. On Blue Gene the acceptable of type are MESH, TORUS and NAV. If NAV, or if not set, then SLURM will try to fit a TORUS else MESH. You should not normally set this option. SLURM will nor- mally allocate a TORUS if possible for a given geometry. -g, --geometry=XxYxZ Specify the geometry requirements for the job. The three numbers represent the required geometry giving dimensions in the X, Y and Z directions. For example "--geometry=2x3x4", specifies a block of nodes having 2 x 3 x 4 = 24 nodes (actually base parti- tions on Blue Gene). --linux-image=path Path to linux image for bluegene block. Default from blu- gene.conf if not set. --mloader-image=path Path to mloader image for bluegene block. Default from blu- gene.conf if not set. -R, --no-rotate Disables rotation of the job's requested geometry in order to fit an appropriate partition. By default the specified geometry can rotate in three dimensions. --ramdisk-image=path Path to ramdisk image for bluegene block. Default from blu- gene.conf if not set. --reboot Force the allocated nodes to reboot before starting the job. srun will submit the job request to the slurm job controller, then ini- tiate all processes on the remote nodes. If the request cannot be met immediately, srun will block until the resources are free to run the job. If the -I (--immediate) option is specified srun will terminate if resources are not immediately available. When initiating remote processes srun will propagate the current work- ing directory, unless --chdir=path is specified, in which case path will become the working directory for the remote processes. The -n, -c, and -N options control how CPUs and nodes will be allo- cated to the job. When specifying only the number of processes to run with -n, a default of one CPU per process is allocated. By specifying the number of CPUs required per task (-c), more than one CPU may be allocated per process. If the number of nodes is specified with -N, srun will attempt to allocate at least the number of nodes specified. Combinations of the above three options may be used to change how pro- cesses are distributed across nodes and cpus. For instance, by specify- ing both the number of processes and number of nodes on which to run, the number of processes per node is implied. However, if the number of CPUs per process is more important then number of processes (-n) and the number of CPUs per process (-c) should be specified. srun will refuse to allocate more than one process per CPU unless --overcommit (-O) is also specified. srun will attempt to meet the above specifications "at a minimum." That is, if 16 nodes are requested for 32 processes, and some nodes do not have 2 CPUs, the allocation of nodes will be increased in order to meet the demand for CPUs. In other words, a minimum of 16 nodes are being requested. However, if 16 nodes are requested for 15 processes, srun will consider this an error, as 15 processes cannot run across 16 nodes. IO Redirection By default, stdout and stderr will be redirected from all tasks to the stdout and stderr of srun, and stdin will be redirected from the stan- dard input of srun to all remote tasks. For OS X, the poll() function does not support stdin, so input from a terminal is not possible. This behavior may be changed with the --output, --error, and --input (-o, -e, -i) options. Valid format specifications for these options are all stdout stderr is redirected from all tasks to srun. stdin is broadcast to all remote tasks. (This is the default behav- ior) none stdout and stderr is not received from any task. stdin is not sent to any task (stdin is closed). taskid stdout and/or stderr are redirected from only the task with relative id equal to taskid, where 0 <= taskid <= ntasks, where ntasks is the total number of tasks in the current job step. stdin is redirected from the stdin of srun to this same task. This file will be written on the node executing the task. filename srun will redirect stdout and/or stderr to the named file from all tasks. stdin will be redirected from the named file and broadcast to all tasks in the job. filename refers to a path on the host that runs srun. Depending on the cluster's file system layout, this may result in the output appearing in different places depending on whether the job is run in batch mode. format string srun allows for a format string to be used to generate the named IO file described above. The following list of format specifiers may be used in the format string to generate a filename that will be unique to a given jobid, stepid, node, or task. In each case, the appropriate number of files are opened and associated with the corresponding tasks. Note that any format string containing %t, %n, and/or %N will be writ- ten on the node executing the task rather than the node where srun executes. %J jobid.stepid of the running job. (e.g. "128.0") %j jobid of the running job. %s stepid of the running job. %N short hostname. This will create a separate IO file per node. %n Node identifier relative to current job (e.g. "0" is the first node of the running job) This will create a separate IO file per node. %t task identifier (rank) relative to current job. This will create a separate IO file per task. A number placed between the percent character and format specifier may be used to zero-pad the result in the IO file- name. This number is ignored if the format specifier corre- sponds to non-numeric data (%N for example). Some examples of how the format string may be used for a 4 task job step with a Job ID of 128 and step id of 0 are included below: job%J.out job128.0.out job%4j.out job0128.out job%j-%2t.out job128-00.out, job128-01.out, ... ENVIRONMENT VARIABLES Some srun options may be set via environment variables. These environ- ment variables, along with their corresponding options, are listed below. Note: Command line options will always override these settings. PMI_FANOUT This is used exclusively with PMI (MPICH2 and MVAPICH2) and controls the fanout of data commu- nications. The srun command sends messages to application programs (via the PMI library) and those applications may be called upon to forward that data to up to this number of additional tasks. Higher values offload work from the srun command to the applications and likely increase the vulnerability to failures. The default value is 32. PMI_FANOUT_OFF_HOST This is used exclusively with PMI (MPICH2 and MVAPICH2) and controls the fanout of data commu- nications. The srun command sends messages to application programs (via the PMI library) and those applications may be called upon to forward that data to additional tasks. By default, srun sends one message per host and one task on that host forwards the data to other tasks on that host up to PMI_FANOUT. If PMI_FANOUT_OFF_HOST is defined, the user task may be required to forward the data to tasks on other hosts. Setting PMI_FANOUT_OFF_HOST may increase performance. Since more work is performed by the PMI library loaded by the user application, failures also can be more common and more difficult to diagnose. PMI_TIME This is used exclusively with PMI (MPICH2 and MVAPICH2) and controls how much the communica- tions from the tasks to the srun are spread out in time in order to avoid overwhelming the srun command with work. The default value is 500 (microseconds) per task. On relatively slow pro- cessors or systems with very large processor counts (and large PMI data sets), higher values may be required. SLURM_CONF The location of the SLURM configuration file. SLURM_ACCOUNT Same as -U, --account=account SLURM_CHECKPOINT Same as --checkpoint=time SLURM_CPU_BIND Same as --cpu_bind=type SLURM_CPUS_PER_TASK Same as -c, --ncpus-per-task=n SLURM_CONN_TYPE Same as --conn-type=(mesh|nav|torus) SLURM_CORE_FORMAT Same as --core=format SLURM_DEBUG Same as -v, --verbose SLURMD_DEBUG Same as -d, --slurmd-debug SLURM_DEPENDENCY -P, --dependency=jobid SLURM_DISABLE_STATUS Same as -X, --disable-status SLURM_DIST_PLANESIZE Same as -m plane SLURM_DISTRIBUTION Same as -m, --distribution=(block|cyclic|arbi- trary) SLURM_EPILOG Same as --epilog=executable SLURM_EXCLUSIVE Same as --exclusive SLURM_GEOMETRY Same as -g, --geometry=X,Y,Z SLURM_JOB_NAME Same as -J, --job-name=jobname except within an existing allocation, in which case it is ignored to avoid using the batch job's name as the name of each job step. SLURM_LABELIO Same as -l, --label SLURM_MEM_BIND Same as --mem_bind=type SLURM_NETWORK Same as --network=type SLURM_NNODES Same as -N, --nodes=(n|min-max) SLURN_NTASKS_PER_CORE Same as --ntasks-per-core SLURN_NTASKS_PER_NODE Same as --ntasks-per-nodea SLURN_NTASKS_PER_SOCKET Same as --ntasks-per-socketa SLURM_NO_ROTATE Same as --no-rotate SLURM_NPROCS Same as -n, --ntasks=n SLURM_OPEN_MODE Same as --open-mode SLURM_OVERCOMMIT Same as -O, --overcommit SLURM_PARTITION Same as -p, --partition=partition SLURM_PROLOG Same as --prolog=executable SLURM_REMOTE_CWD Same as -D, --chdir==dir SLURM_SRUN_COMM_IFHN Same as --ctrl-comm-ifhn=addr SLURM_STDERRMODE Same as -e, --error=mode SLURM_STDINMODE Same as -i, --input=mode SLURM_STDOUTMODE Same as -o, --output=mode SLURM_TASK_EPILOG Same as --task-epilog=executable SLURM_TASK_MEM Same as --task-mem SLURM_TASK_PROLOG Same as --task-prolog=executable SLURM_THREADS Same as -T, --threads SLURM_TIMELIMIT Same as -t, --time=minutes SLURM_UNBUFFEREDIO Same as -u, --unbuffered SLURM_WAIT Same as -W, --wait=seconds SLURM_WORKING_DIR -D, --chdir=path Additionally, srun will set some environment variables in the environ- ment of the executing tasks on the remote compute nodes. These envi- ronment variables are: SLURM_CPU_BIND_VERBOSE --cpu_bind verbosity (quiet,verbose). SLURM_CPU_BIND_TYPE --cpu_bind type (none,rank,map_cpu:,mask_cpu:) SLURM_CPU_BIND_LIST --cpu_bind map or mask list () SLURM_CPUS_ON_NODE Count of processors available to the job on this node. Note the select/linear plugin allocates entire nodes to jobs, so the value indicates the total count of CPUs on the node. The select/cons_res plugin allocates individual processors to jobs, so this number indicates the number of processors on this node allocated to the job. SLURM_GTIDS Global task IDs running on this node. Zero origin and comma separated. SLURM_JOB_DEPENDENCY Set to value of the --dependency option. SLURM_JOBID Job id of the executing job SLURM_LAUNCH_NODE_IPADDR IP address of the node from which the task launch was initiated (where the srun command ran from) SLURM_LOCALID Node local task ID for the process within a job SLURM_MEM_BIND_VERBOSE --mem_bind verbosity (quiet,verbose). SLURM_MEM_BIND_TYPE --mem_bind type (none,rank,map_mem:,mask_mem:) SLURM_MEM_BIND_LIST --mem_bind map or mask list () SLURM_NNODES Total number of nodes in the job's resource allocation SLURM_NODEID The relative node ID of the current node SLURM_NODELIST List of nodes allocated to the job SLURM_NPROCS Total number of processes in the current job SLURM_PRIO_PROCESS The scheduling priority (nice value) at the time of job submis- sion. This value is propagated to the spawned processes. SLURM_PROCID The MPI rank (or relative process ID) of the current process SLURM_STEPID The step ID of the current job SLURM_TASK_PID The process ID of the task being started. SLURM_TASKS_PER_NODE Number of tasks to be initiated on each node. Values are comma separated and in the same order as SLURM_NODELIST. If two or more consecutive nodes are to have the same task count, that count is followed by "(x#)" where "#" is the repetition count. For example, "SLURM_TASKS_PER_NODE=2(x3),1" indicates that the first three nodes will each execute three tasks and the fourth node will execute one task. SLURM_UMASK The umask (user file-create mask) at the time of job submission. This value is propagated to the spawned processes. MPIRUN_NOALLOCATE Do not allocate a block on Blue Gene systems only. MPIRUN_NOFREE Do not free a block on Blue Gene systems only. MPIRUN_PARTITION The block name on Blue Gene systems only. SIGNALS AND ESCAPE SEQUENCES Signals sent to the srun command are automatically forwarded to the tasks it is controlling with a few exceptions. The escape sequence will report the state of all tasks associated with the srun command. If is entered twice within one second, then the associated SIGINT signal will be sent to all tasks and a termination sequence will be entered sending SIGCONT, SIGTERM, and SIGKILL to all spawned tasks. If a third is received, the srun program will be terminated without waiting for remote tasks to exit or their I/O to complete. The escape sequence is presently ignored. Our intent is for this put the srun command into a mode where various special actions may be invoked. MPI SUPPORT MPI use depends upon the type of MPI being used. There are three fun- damentally different modes of operation used by these various MPI implementation. 1. SLURM directly launches the tasks and performs initialization of communications (Quadrics MPI, MPICH2, MPICH-GM, MVAPICH, MVAPICH2 and some MPICH1 modes). For example: "srun -n16 a.out". 2. SLURM creates a resource allocation for the job and then mpirun launches tasks using SLURM's infrastructure (OpenMPI, LAM/MPI, HP-MPI and some MPICH1 modes). 3. SLURM creates a resource allocation for the job and then mpirun launches tasks using some mechanism other than SLURM, such as SSH or RSH (BlueGene MPI and some MPICH1 modes). These tasks initiated out- side of SLURM's monitoring or control. SLURM's epilog should be config- ured to purge these tasks when the job's allocation is relinquished. See https://computing.llnl.gov/linux/slurm/quickstart.html#mpi for more information on use of these various MPI implementation with SLURM. MULTIPLE PROGRAM CONFIGURATION Comments in the configuration file must have a "#" in column one. The configuration file contains the following fields separated by white space: Task rank One or more task ranks to use this configuration. Multiple val- ues may be comma separated. Ranges may be indicated with two numbers separated with a '-' with the smaller number first (e.g. "0-4" and not "4-0"). To indicate all tasks, specify a rank of '*' (in which case you probably should not be using this option). If an attempt is made to initiate a task for which no executable program is defined, the following error message will be produced "No executable program specified for this task". Executable The name of the program to execute. May be fully qualified pathname if desired. Arguments Program arguments. The expression "%t" will be replaced with the task's number. The expression "%o" will be replaced with the task's offset within this range (e.g. a configured task rank value of "1-5" would have offset values of "0-4"). Single quotes may be used to avoid having the enclosed values inter- preted. This field is optional. For example: ################################################################### # srun multiple program configuration file # # srun -n8 -l --multi-prog silly.conf ################################################################### 4-6 hostname 1,7 echo task:%t 0,2-3 echo offset:%o > srun -n8 -l --multi-prog silly.conf 0: offset:0 1: task:1 2: offset:1 3: offset:2 4: linux15.llnl.gov 5: linux16.llnl.gov 6: linux17.llnl.gov 7: task:7 EXAMPLES This simple example demonstrates the execution of the command hostname in eight tasks. At least eight processors will be allocated to the job (the same as the task count) on however many nodes are required to sat- isfy the request. The output of each task will be proceeded with its task number. (The machine "dev" in the example below has a total of two CPUs per node) > srun -n8 -l hostname 0: dev0 1: dev0 2: dev1 3: dev1 4: dev2 5: dev2 6: dev3 7: dev3 The output of test.sh would be found in the default output file "slurm-42.out." The srun -r option is used within a job script to run two job steps on disjoint nodes in the following example. The script is run using allo- cate mode instead of as a batch job in this case. > cat test.sh #!/bin/sh echo $SLURM_NODELIST srun -lN2 -r2 hostname srun -lN2 hostname > salloc -N4 test.sh dev[7-10] 0: dev9 1: dev10 0: dev7 1: dev8 The follwing script runs two job steps in parallel within an allocated set of nodes. > cat test.sh #!/bin/bash srun -lN2 -n4 -r 2 sleep 60 & srun -lN2 -r 0 sleep 60 & sleep 1 squeue squeue -s wait > salloc -N4 test.sh JOBID PARTITION NAME USER ST TIME NODES NODELIST 65641 batch test.sh grondo R 0:01 4 dev[7-10] STEPID PARTITION USER TIME NODELIST 65641.0 batch grondo 0:01 dev[7-8] 65641.1 batch grondo 0:01 dev[9-10] This example demonstrates how one executes a simple MPICH job. We use srun to build a list of machines (nodes) to be used by mpirun in its required format. A sample command line and the script to be executed follow. > cat test.sh #!/bin/sh MACHINEFILE="nodes.$SLURM_JOBID" # Generate Machinefile for mpich such that hosts are in the same # order as if run via srun # srun -l /bin/hostname | sort -n | awk '{print $2}' > $MACHINEFILE # Run using generated Machine file: mpirun -np $SLURM_NPROCS -machinefile $MACHINEFILE mpi-app rm $MACHINEFILE > salloc -N2 -n4 test.sh This simple example demonstrates the execution of different jobs on different nodes in the same srun. You can do this for any number of nodes or any number of jobs. The executables are placed on the nodes sited by the SLURM_NODEID env var. Starting at 0 and going to the num- ber specified on the srun commandline. > cat test.sh case $SLURM_NODEID in 0) echo "I am running on " hostname ;; 1) hostname echo "is where I am running" ;; esac > srun -N2 test.sh dev0 is where I am running I am running on dev1 This example demonstrates use of multi-core options to control layout of tasks. We request that four sockets per node and two cores per socket be dedicated to the job. > srun -N2 -B 4-4:2-2 a.out This example shows a script in which Slurm is used to provide resource management for a job by executing the various job steps as processors become available for their dedicated use. > cat my.script #!/bin/bash srun --exclusive -n4 prog1 & srun --exclusive -n3 prog2 & srun --exclusive -n1 prog3 & srun --exclusive -n1 prog4 & wait COPYING Copyright (C) 2006-2007 The Regents of the University of California. Copyright (C) 2008 Lawrence Livermore National Security. Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER). LLNL-CODE-402394. This file is part of SLURM, a resource management program. For details, see . 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. SEE ALSO salloc(1), sttach(1), sbatch(1), sbcast(1), scancel(1), scontrol(1), squeue(1), slurm.conf(5), sched_setaffinity(2), numa(3) getrlimit(2), srun 1.3 August 2008 SRUN(1)