Job submit/management commands
Job submit command
Job submit command (qsub command)
Submit command places a job on the queue of the accelerator server.
Submitting options also can be described in the execution script.
For more details, see each manual.
attention!
Please submit job to appropriate queue with qsub command, not to execute your program on the front end node directly.If you execute a program on the front end node, it may be canceled by the administrator because it has effects on other users.
Usage
$ qsub [-q queue name] [-l select=number of nodes] [-N Job name] [-M e-mail address] [-m mail point] [-l walltime=limit of walltime] [-l license name=number of use licenses] [execution script file]
List of options
| Option | Value |
|---|---|
| q queue name | Set the queue name. As for details of queue, see List of queues. |
| -l select=number of nodes | Specify number of the nodes to be used. No this option means default (see List of queues). |
| -N job name | Set the job name. The name of job is up to 236 characters. The real-time job reference system displays up to 64 characters. If no ‘-N’ option is specified, the system assign the default job name. |
| -M e-mail address | Set e-mail address to be received. To receive an e-mail, ‘-m’ option is required. |
| -m mail point | Specify the set of events that causes mailing to be sent to the list of users specified in the ‘-M’ option. To receive an e-mail, ‘-m’ option is required. |
| -l walltime=limit of walltime | Specify the limit of walltime. If no this option is specified, the walltime is default value specified in the queue. (see List of queues) Appropriate value makes a queued job running frequently. |
| -l license name=number of use licenses | Specify the number of licenses when you use applications that requires managed licenses. If no this options specified, the job is regarded not to use applications that licenses are managed. As for specifying licenses, see how to execute applications. |
Example
Execute following jobs
・queue: A_004, number of nodes: 2, limit of walltime: 1 hour, script file: hello.sh
$ qsub -q A_004 -l select=2 -l walltime=1:00:00 hello.sh
Description in the execution script.
#!/bin/sh #PBS -q A_004 #PBS -l select=2 #PBS -l walltime=1:00:00 : : :
・queue: DA_002g, script file: hello.sh, recipient address: userA@test.com
Mail is sent when the job begins execution (option is ‘-m b’) and terminates (option is ‘-m e’).
$ qsub -q DA_002g -M userA@test.com -m be hello.sh
Description in the execution script.
#!/bin/sh #PBS -q DA_002g #PBS -M userA@test.com #PBS -m be : : :
・queue: A_004, job name: TEST, script file: hello.sh
$ qsub -q A_004 -N TEST hello.sh
Description in the execution script.
#!/bin/sh #PBS -q A_004 #PBS -N TEST : : :
Format of an execution script
This section describes the format of execution script files to run programs on the accelerator server.
To execute the application that requires an execution script file, create the file in advance.
/work area have better I/O performance than /home area, so copy your data to the /work area, execute it, and move the result to the source directory with reference to the following example.
For more details, see each manual.
Execute a non MPI program
#!/bin/sh #PBS -l select=1 #PBS -q queue #PBS -N jobname # Copy the job input directory to /work area and move there DIRNAME=`basename $PBS_O_WORKDIR` WORKDIR=/work/$USER/$PBS_JOBID mkdir -p $WORKDIR cp -raf $PBS_O_WORKDIR $WORKDIR cd $WORKDIR/$DIRNAME # Execute your program program > output file 2> error file # Move the result to the source directory after execution cd; if cp -raf $WORKDIR/$DIRNAME $PBS_O_WORKDIR/.. ; then rm -rf $WORKDIR; fi
・Example to execute a program ‘a.out’.
#!/bin/sh #PBS -l select=1 #PBS -q A_004 #PBS -N sample DIRNAME=`basename $PBS_O_WORKDIR` WORKDIR=/work/$USER/$PBS_JOBID mkdir -p $WORKDIR cp -raf $PBS_O_WORKDIR $WORKDIR cd $WORKDIR/$DIRNAME ./a.out > result.out 2> result.err cd; if cp -raf $WORKDIR/$DIRNAME $PBS_O_WORKDIR/.. ; then rm -rf $WORKDIR; fi
Execute a MPI program
#!/bin/sh #PBS -l select=nodes #PBS -q queue #PBS -N jobname # Copy the job input directory to /work area and move there DIRNAME=`basename $PBS_O_WORKDIR` WORKDIR=/work/$USER/$PBS_JOBID mkdir -p $WORKDIR cp -raf $PBS_O_WORKDIR $WORKDIR cd $WORKDIR/$DIRNAME # Execute your program mpirun [ -np MPI total tasks | -ppn MPI tasks per node ] -hostfile $PBS_NODEFILE program > output file 2> error file # Move the result to the source directory after execution cd; if cp -raf $WORKDIR/$DIRNAME $PBS_O_WORKDIR/.. ; then rm -rf $WORKDIR; fi
・Example to run a program on 2 node and 72 MPI processes using Intel compiler.
#!/bin/sh #PBS -l select=2 #PBS -q A_004 #PBS -N mpi DIRNAME=`basename $PBS_O_WORKDIR` WORKDIR=/work/$USER/$PBS_JOBID mkdir -p $WORKDIR cp -raf $PBS_O_WORKDIR $WORKDIR cd $WORKDIR/$DIRNAME mpirun -np 72 -ppn 36 -hostfile $PBS_NODEFILE ./a.out > result.out 2> result.err cd; if cp -raf $WORKDIR/$DIRNAME $PBS_O_WORKDIR/.. ; then rm -rf $WORKDIR; fi
Interactive mode
Submit a job with interactive mode.
Add option -I (uppercase i) to the qsub command and specify IA_001g, CA_001, CA_001g, IC_001 or CC_001 for the queue.
Usage
$ qsub -I -q queue
Example
$ qsub -I -q IA_001g qsub: waiting for job 22351.gpu1 to start qsub: job 22351.gpu1 ready -bash-4.2$./a.out
Submitting jobs to Shared-Queue CA_001, CA_001g and CC_001
CA_001, CA_001g and CC_001 are a queue used by sharing the node with other jobs.
Jobs executed on CA_001 or CA_001g are assigned 1CPU and 1GPU by default, up to 18 CPU and 5 GPU. Execution in interactive mode is also available.
Jobs executed on CC_001 are assigned 1CPU by default, up to 18 CPU. Execution in interactive mode is also available.
Usage
・CA_001 or CA_001g
$ qsub -q queue [ -I ] [ -l select=1[:ncpus= number of CPU][:ngpus= number of GPU] [execution script file] ]
・CC_001
$ qsub -q queue [ -I ] [ -l select=1[:ncpus= number of CPU] [execution script file] ]
Example
・queue: CA_001, Command to execute interactive mode with 2 CPU and 1 GPU
$ qsub -I -q CA_001 -l select=1:ncpus=2:ngpus=1 qsub: waiting for job 22351.gpu1 to start qsub: job 22351.gpu1 ready -bash-4.2$ ./a.out
・queue: CA_001g, Script to execute program a.out with 18 CPU and 5 GPU
To use the queue CA_001g, please submit a job from /work_da area.
#!/bin/sh #PBS -l select=1:ncpus=18:ngpus=5 #PBS -q CA_001g #PBS -N sample DIRNAME=`basename $PBS_O_WORKDIR` WORKDIR=/work/$USER/$PBS_JOBID mkdir -p $WORKDIR cp -raf $PBS_O_WORKDIR $WORKDIR cd $WORKDIR/$DIRNAME mpirun -np 18 -ppn 18 -hostfile $PBS_NODEFILE ./a.out > result.out 2> result.err cd; if cp -raf $WORKDIR/$DIRNAME $PBS_O_WORKDIR/.. ; then rm -rf $WORKDIR; fi
・ queue: CC_001, Command to execute interactive mode with 2 CPU
$ qsub -I -q CC_001 -l select=1:ncpus=2 qsub: waiting for job 90289.gpu1 to start qsub: job 90289.gpu1 ready -bash-4.2$./a.out
・ queue: CC_001, Script to execute program a.out with 18 CPU
#!/bin/sh #PBS -l select=1:ncpus=18 #PBS -q CC_001 #PBS -N sample DIRNAME=`basename $PBS_O_WORKDIR` WORKDIR=/work/$USER/$PBS_JOBID mkdir -p $WORKDIR cp -raf $PBS_O_WORKDIR $WORKDIR cd $WORKDIR/$DIRNAME mpirun -np 18 -ppn 18 -hostfile $PBS_NODEFILE ./a.out > result.out 2> result.err cd; if cp -raf $WORKDIR/$DIRNAME $PBS_O_WORKDIR/.. ; then rm -rf $WORKDIR; fi
Job management command
Display your own job information
Description
Display your own job information on the supercomputer.
Usage
$ statj [-x] [ [job_identifier | destination] ...]
List of options
| Option | Value |
|---|---|
| -x | Displays status of finished, queued, and running jobs. |
Example
userA@gpu2:~> statj
Req'd Req'd Elap
Job ID Username Queue Jobname SessID NDS TSK Memory Time S Time
----------------- --------------- --------- ----------------
3413.gpu1 userA A_004 STDIN 231503 1 36 690gb 24:00 R 00:00
Display job information
Description
Display information of the jobs on the accelerator server.
Usage
Default format: qstat [-a] [-p] [-J] [-t] [-x] [ [job_identifier | destination] ...] Long format: qstat -f [-p] [-J] [-t] [-x] [ [job_identifier | destination] ...]
List of options
| Option | Value |
|---|---|
| -a | Display memory usage, elapsed time, status of jobs, etc. |
| -p | Display the percentage of the job completion. |
| -J | Display limits status of job array. |
| -t | Displays status of jobs. |
| -x | Displays status of finished, queued, and running jobs. |
| -f | Display status in long format. |
Example
userA@gpu2:~> qstat -a
Req'd Req'd Elap
Job ID Username Queue Jobname SessID NDS TSK Memory Time S Time
--------------- -------- -------- ---------- ------ --- --- ------ ----- - -----
3390.gpu1 userA A_004 abinit 193347 4 144 2760gb 72:00 R 47:28
3401.gpu1 userA A_004 prog9_1 121974 4 144 2760gb 72:00 R 47:26
userA@gpu2:~> qstat -p
Job id Name User % done S Queue
---------------- ---------------- ---------------- -------- - -----
3390.gpu1 abinit userA 2 R A_004
3401.gpu1 prog9_1 userA 0 R A_004
userA@gpu2:~> qstat -t
Job id Name User Time Use S Queue
---------------- ---------------- ---------------- -------- - -----
3390.gpu1 abinit userA 00:00:01 R A_004
3401.gpu1 prog9_1 userA 00:00:01 R A_004
userA@gpu2:~ > qstat -x
Job id Name User Time Use S Queue
---------------- ---------------- ---------------- -------- - -----
2235.gpu1 prog9_2 userA 00:00:03 F A_016
2236.gpu1 vasp4 userA 00:00:01 F A_016
2237.gpu1 prog9_1 userA 00:00:01 F A_016
The rest is omitted
...
userA@gpu2:~> qstat -f 3390.gpu1
Job Id: 3390.gpu1
Job_Name = abinit
Job_Owner = userA@gpu2
resources_used.cpupercent = 10
resources_used.cput = 00:00:01
resources_used.mem = 12836kb
resources_used.ncpus = 72
The rest is omitted
...
Display queue status
Description
Display information about queues on the accelerator server.
Usage
Default format: statq [destination ...] Long format: statq -f [destination ...]
List of options
| Option | Value |
|---|---|
| -f | Display status in long format. |
Example
userA@gpu2:~> statq
Queue Max Tot Ena Str Que Run Hld Wat Trn Ext Type
---------------- ----- ----- --- --- ----- ----- ----- ----- ----- ----- ----
workq 0 0 no yes 0 0 0 0 0 0 Exec
A_004 0 0 yes yes 0 0 0 0 0 0 Exec
A_008 0 1 yes yes 0 1 0 0 0 0 Exec
A_016 0 0 yes yes 0 0 0 0 0 0 Exec
DA_002g 0 0 yes yes 0 0 0 0 0 0 Exec
DC_002 0 0 yes yes 0 0 0 0 0 0 Exec
C_002 0 0 yes yes 0 0 0 0 0 0 Exec
The rest is omitted
...
userA@gpu2:~> statq -f
Queue: workq
queue_type = Execution
total_jobs = 0
state_count = Transit:0 Queued:0 Held:0 Waiting:0 Running:0 Exiting:0 Begun:0
enabled = False
started = True
The rest is omitted
...
Display server status
Description
Display information about servers of the accelerator server.
Usage
Default format: qstat -B [destination ...] Long format: qstat -B -f [destination ...]
List of options
| Option | Value |
|---|---|
| -B | Display server status. |
| -f | Display status in long format. |
Example
userA@gpu2:~> qstat -B
Server Max Tot Que Run Hld Wat Trn Ext Status
---------------- ----- ----- ----- ----- ----- ----- ----- ----- -----------
gpu1 0 1155 0 1 0 0 0 0 Active
userA@gpu2:~> qstat -Bf
Server: sdb
server_state = Active
server_host = sdb
scheduling = True
max_queued = [u:PBS_GENERIC=200]
The rest is omitted
...
Cancel the job before the job finished
Description
Use qdel command for cancelling the job on the accelerator server.
Usage
qdel [ -x ] [ -Wsuppress_email=<N> ] job_identifier [job_identifier ...]
List of options
| Option | Value |
|---|---|
| -x | Delete job and job history of the specified job. |
| -Wsuppress_email | Set limit on number of e-mails sent on deleting jobs. |
Example
userA@gpu2:~/work/20180712_sample> qstat Job id Name User Time Use S Queue ---------------- ---------------- ---------------- -------- - ----- 3413.gpu1 abinit userA 00:00:00 R A_004 3414.gpu1 STDIN_gpu2_22 userA 00:00:00 R A_004 userA@gpu2:~/work/20180712_sample> qdel 3414.gpu1 userA@gpu2:~/work/20180712_sample> qstat Job id Name User Time Use S Queue ---------------- ---------------- ---------------- -------- - ----- 3413.gpu1 abinit userA 00:00:00 R A_004 userA@gpu2:~/work/20180712_sample>
Display information about used and remained time of job execution (jobtime command)
Description
Use jobtime command for listing information of your completed jobs.
Usage
$ jobtime
Information
| Section | Details |
|---|---|
| Last Updated | updated time |
| User | user ID |
| Total | available time of job execution |
| Used | used time |
| Remained | remained time |
Example
userA@gpu2:~ > jobtime # Last Updated: 2018/10/01 13:45 # User Total Used Remained (H) username 500 222.32 277.68
References for submitting job and script
How to execute a MPI job
Description
Intel MPI is available for MPI environment.
Usage
Use mpirun command to execute a job.
mpirun [ -np parallel number ] [ -ppn parallel number per node ] -hostfile $PBS_NODEFILE execution program
important
To improve job performance, specify values so that:[ Parallel number(the value of “-np”) ] = [ Number of nodes(the value of “#PBS -l select=”) ] × [ Parallel number per node(the value of “-ppn”) ]
Example
#!/bin/bash #PBS -j oe #PBS -l select=1 DIRNAME=`basename $PBS_O_WORKDIR` WORKDIR=/work/$USER/$PBS_JOBID mkdir -p $WORKDIR cp -raf $PBS_O_WORKDIR $WORKDIR cd $WORKDIR/$DIRNAME mpirun -np 36 -hostfile $PBS_NODEFILE /usr/local/app/ABINIT/current/src/98_main/abinit < input.files > result.out 2> result.err cd; if cp -raf $WORKDIR/$DIRNAME $PBS_O_WORKDIR/.. ; then rm -rf $WORKDIR; fi