While you are reading this document we assume that you are able to Enroll for NSCC services and able to login to NSCC cluster using one of the login methods

Document Conventions

Command line

@ node where you execute the commands
#	execute commands --------------------------------------------------------------------------------------------------------------- the result of commands.

@ node where you execute the commands

execute commands

---------------------------------------------------------------------------------------------------------------

the result of commands.

↑

# execute as a root

$ execute as a general user

Configuration file

Submission file/script
Contents of the file. red line : need to be changed

Submission file/script

Contents of the file.

red line : need to be changed

Support

To receive technical support from NSCC System administrator, please do write us [email protected] or raise a ticket using https://help.nscc.sg ->User Services -> Service Desk Portal

Introduction to NSCC Job Scheduler (PBS Pro)

The jobs scheduler installed and configured for use on theNSCC Supercomputer is PBS Pro.

It provides a workload management solution that maximizes the efficiency and utilization of high-performance computing (HPC) resources and improves job turnaround.

Some of the important features of PBS Pro are below

Robust Workload Management

Floating licenses
Scalability, with flexible queues
Job arrays
User and administrator interface
Job suspend/resume
Application checkpoint/restart
Automatic file staging
Accounting logs
Access control lists

Advanced Scheduling Algorithms

Resource-based scheduling
Preemptive scheduling
Optimized node sorting
Enhanced job placement
Advance & standing reservations
Cycle harvesting across workstations
Scheduling across multiple complexes
Network topology scheduling
Manages both batch and interactive work
Backfilling

Reliability, Availability and Scalability

Server failover feature
Automatic job recovery
System monitoring
Integration with MPI solutions
Tested to manage 1,000,000+ jobs per day
Tested to accept 30,000 Jobs per minute
EAL3+ security
Checkpoint support

Getting Started with PBS Pro

NSCC is using PBS Pro to schedule jobs on the cluster. This guide contains the basic information needed for users to start submitting jobs.

PBS Queues

Several queues have been created to satisfy the resource requirements of the various workloads which use the system. The subsequent sections provide an overview of the queue definitions and limits.

External Queues:

End users must submit their jobs to one of the external queues.

Queue Name	Resources Available	Remarks
normal	12 Compute nodes 24 cores per server 4GB/core memory or 96 GB per server	The queue normal is a routing queue and does not execute any jobs. It only routes the job to the internal queues based on the resource requirement
gpu	128 GPU nodes 24 Cores per server 4GB/core memory or 96 GB memory per server	This queue is a routing queue and does not execute any jobs. This queue routes the jobs to the internal queues based on resource requirement
largemem	9 Compute nodes 24/48 cores per server 1TB/4TB/6TB memory configurations	This is an execution queue which can take large memory jobs which requires more than 96GB of memory per server
production	480 compute cores reserved for GIS	Only selected GIS users can submit jobs to this queue
provision	Special Administrative queue	No access to users
iworkq	Special queue for visualization jobs	Jobs submitted from Display manager to be dispatched to this queue
imeq	Special queue for ime jobs	This queue is under testing

Internal Queues:

These queues are typically execution queues which actually schedule the jobs based on the resources that they request

These queues are reference purpose only and cannot accept jobs directly from users.

Queue Name	Queue Limitations	Run time	Remarks
long	One node per user at a time	120 Hrs	Queue for the jobs which cannot be checkpointed and cannot scale up more than a single node. Longer queue time
dev	Up to 48 cores	1 Hr	Designed to test application execution and jobs can be dispatched almost immediately
q1	Upto 23 cores	24 Hrs	Small jobs which require maximum of 1 compute node. Less queue time
medium	Minimum of 24 cores and maximum 599 cores	24 Hrs	Jobs that can scale more than 1 compute node up to 599 cores
q4	Minimum of 600 cores	24 Hrs	Big jobs that can scale well with more than 600 CPU cores and less queueing time
gpunormal	Minimum 1 GPU and maximum 128 GPUs	24 Hrs	Queue for normal GPU jobs which can request for 1 or more GPUs
gpulong	Minimum 1 GPU and maximum 128 GPUs	240 Hrs	Queue for long GPU jobs which cannot be checkpointed. Longer queue time expected when requested more GPU nodes

The above mentioned values/limitations may change without any prior notice

System wide restrictions

Below are few System wide restrictions that apply for job submission

Job submission to Internal Queues : Denied by Policy and System

Executing applications in Login nodes : Denied by Policy

Per core Memory : 4GB/core applied by System

Cores per server/node : 24 Cores per server applied by System

Cluster wide cores : 3000 Cores per user

Cluster wide No of Jobs : 300 Jobs in Queue with all states

Job Submission

The job submission in NSCC will be illustrated through few examples:

Example 1: Simple Job submission

Login node
$	echo sleep 10 \|qsub -q normal -l select=1:ncpus=1,walltime=00:13:00 --------------------------------------------------------------------------------------------------------------- 154400.wlm01

echo sleep 10 |qsub -q normal -l select=1:ncpus=1,walltime=00:13:00

---------------------------------------------------------------------------------------------------------------

154400.wlm01

In the above example, the job will be submitted with one command line: sleep 10

Example 2: Submit job using job submission script

Login node
$	qsub -q normal -l select=1:ncpus=1,walltime=00:00:40 –P Personal submit.pbs ------------------------------------------------------------------- 154400.wlm01

qsub -q normal -l select=1:ncpus=1,walltime=00:00:40 –P Personal submit.pbs

-------------------------------------------------------------------

154400.wlm01

In the above command, walltime can be changed according to requirement. The directive -P Personal can be changed to -P <Project ID>.

Contents of submit.pbs are as below

Login node
$	cat submit.pbs ------------------------------------------------------------------- #!/bin/bash #PBS –N Sleep Job #PBS –j oe echo sleep job for 30 seconds sleep 30

cat submit.pbs

-------------------------------------------------------------------

#!/bin/bash

#PBS –N Sleep Job

#PBS –j oe

echo sleep job for 30 seconds

sleep 30

In the above example, resource requested for the job using command line and rest of the job execution commands are in the job submission script. Each of the job script commands are explained below.

#PBS is the pbs directive

-q normal -> Choose the queue “normal”

-l select=1:ncpus=1 -> reserves 1 CPU core for this job

#PBS –N Sleep Job

-> -N states Name of the job

-> Sleep Job is the name of my job

#PBS -j oe

-> -j states to join output and error files together

Example 3: Submit job using job submission script, all options in the job submission script

Login node
$	qsub submit.pbs ------------------------------------------------------------------- 154400.wlm01

qsub submit.pbs

-------------------------------------------------------------------

154400.wlm01

Contents of submit.pbs are below

Login node
$	cat submit.pbs ------------------------------------------------------------------- #!/bin/bash #PBS –q normal #PBS –l select=1:ncpus=1:mem=100M #PBS –l walltime=00:10:00 #PBS –P Personal # can be replace with the project ID #PBS –N Sleep_Job #PBS -o ~/outputfiles/Sleep_Job.o #PBS –e ~/errorfiles/Sleep_Job.e echo sleep job for 30 seconds sleep 30

cat submit.pbs

-------------------------------------------------------------------

#!/bin/bash

#PBS –q normal

#PBS –l select=1:ncpus=1:mem=100M

#PBS –l walltime=00:10:00

#PBS –P Personal # can be replace with the project ID

#PBS –N Sleep_Job

#PBS -o ~/outputfiles/Sleep_Job.o

#PBS –e ~/errorfiles/Sleep_Job.e

echo sleep job for 30 seconds

sleep 30

In the above example, the resources are requested using job script. Below is the explanation of each line.

#PBS is the pbs directive

#PBS -q normal

-> Choose the queue “normal”

#PBS -l select=1:ncpus=1

-> reserves 1 cpus for this job

#PBS –l walltime=00:10:00

-> Request for a wall time of 10 Minutes

#PBS –P Personal

-> Request PBS to account under Personal project

#PBS –N Sleep Job

-> -N states Name of the job

-> Sleep Job is the name of my job

#PBS -o ~/outputfiles/Sleep_Job.o

-> -o states to create output file “Sleep_Job.o” under the directory ~/outputfiles

#PBS -e ~/errorfiles/Sleep_Job.e

-> -e states to create error file “Sleep_Job.e” under the directory ~/errorfiles

Example4: Submit job for a pre-compiled application for demonstration purpose application R is used for serial job.

Login node
$	cat submit.pbs ------------------------------------------------------------------- #!/bin/bash #PBS –q normal #PBS –l select=1:ncpus=1:mem=1G #PBS –l walltime=01:00:00 #PBS –P Personal #PBS –N R_Sample cd $PBS_O_WORKDIR module load R/3.2.4 R CMD BATCH simple.R # Output of R is generated to simple.Rout

cat submit.pbs

-------------------------------------------------------------------

#!/bin/bash

#PBS –q normal

#PBS –l select=1:ncpus=1:mem=1G

#PBS –l walltime=01:00:00

#PBS –P Personal

#PBS –N R_Sample

cd $PBS_O_WORKDIR

module load R/3.2.4

R CMD BATCH simple.R

# Output of R is generated to simple.Rout

In the above example, the resources required for the job are requested using job script. Below is the explanation of each line.

#PBS is the pbs directive

#PBS -q normal

-> Choose the queue “normal”

#PBS -l select=1:ncpus=1

-> reserves 1 unit of 1 CPU core and 1Gb of memory for this job

#PBS –l walltime=01:00:00

-> Request a wall time of 1 Hour

#PBS –P Personal

-> Request PBS to account under Personal project

#PBS –N R_Sample

-> -N states Name of the job

-> R_sample is the name of my job

cd $PBS_O_WORKDIR

-> Execute following commands in the current working directory

module load R/3.2.4

-> Load the environment for R/3.2.4

R CMD BATCH simple.R

-> R execution command

* To submit the job use the qsub syntax in example 3

Example5: Submit job for a MPI application for demonstration purpose gromacs application is used.

Login node
$	cat submit.pbs ------------------------------------------------------------------- #!/bin/bash #PBS –q normal #PBS –l select=2:ncpus=24:mem=96G:mpiprocs=24:ompthreads=1 #PBS –l walltime=16:00:00 #PBS –P Personal #PBS –N gromacs_example module load gromacs/5.1.2/gcc493/impi cd $PBS_O_WORKDIR mpirun mdrun_mpi -pin on -ntomp $OMP_NUM_THREADS -s inputfile.tpr > gromacs.log

cat submit.pbs

-------------------------------------------------------------------

#!/bin/bash

#PBS –q normal

#PBS –l select=2:ncpus=24:mem=96G:mpiprocs=24:ompthreads=1

#PBS –l walltime=16:00:00

#PBS –P Personal

#PBS –N gromacs_example

module load gromacs/5.1.2/gcc493/impi

cd $PBS_O_WORKDIR

mpirun mdrun_mpi -pin on -ntomp $OMP_NUM_THREADS -s inputfile.tpr > gromacs.log

In the above script the resources are requested through the job script. Each line of the job script is explained below.

#PBS is the pbs directive

#PBS -q normal

-> Choose the queue “normal”

#PBS –l select=2:ncpus=24:mem=96G:mpiprocs=24:ompthreads=1

-> reserves 2 units of 24 cpu, 96Gb memory, 24 MPI processes and one OpenMP thread for this job

#PBS –l walltime=16:00:00

-> Request a wall time of 16 Hour

#PBS –N gromacs_example

-> Name of the job defined as gromacs_example

#PBS –P Personal

-> Request PBS to account under Personal project

module load gromacs/5.1.2/gcc493/impi

-> Loads the environment for gromacs/5.1.2/gcc493/impicd $PBS_O_WORKDIR

-> Execute commands in current working directory

mpirun mdrun_mpi -pin on -ntomp $OMP_NUM_THREADS -s inputfile.tpr >gromacs.log

-> gromacs execution command with mpirun

* To submit the job use the qsub syntax in example 3

* In this command notice mpirun does not have any arguments as PBS Pro is tightly integrated with mpirun command

* In case of application enabled with openmp and MPI, it is suggested to use mpiprocs times ompthreads less than or equals to 24.

Example5: Submit job for a MPI application for gpu queue. For demonstration purpose gromacs application is used.

Login node
$	cat submit.pbs ------------------------------------------------------------------- #!/bin/bash #PBS –q gpu #PBS –l select=2:ncpus=24:mem=96G:ngpus=1:mpiprocs=2:ompthreads=12 #PBS –l walltime=16:00:00 #PBS –P Personal #PBS –N gromacs_example module load gromacs/5.1.2/gcc493/impi_cuda cd $PBS_O_WORKDIR mpirun mdrun_mpi -pin on -ntomp $OMP_NUM_THREADS -s inputfile.tpr >mdrun.log

cat submit.pbs

-------------------------------------------------------------------

#!/bin/bash

#PBS –q gpu

#PBS –l select=2:ncpus=24:mem=96G:ngpus=1:mpiprocs=2:ompthreads=12

#PBS –l walltime=16:00:00

#PBS –P Personal

#PBS –N gromacs_example

module load gromacs/5.1.2/gcc493/impi_cuda

cd $PBS_O_WORKDIR

mpirun mdrun_mpi -pin on -ntomp $OMP_NUM_THREADS -s inputfile.tpr >mdrun.log

In the above script the resources are requested through the job script. Each line of the job script is explained below.

#PBS is the pbs directive

#PBS –q gpu

-> requesting for gpu queue

#PBS –l select=2:ncpus=24:mem=96G:ngpus=1:mpiprocs=2:ompthreads=12

-> Request 2 units of 24 CPUs, 96GB of memory, 1 gpu, and 12 threads with 2 MPI Processes

#PBS –l walltime=16:00:00

-> Request for a wall time of 16 Hours

#PBS –P Personal

-> Request PBS to account under Personal project

#PBS –N gromacs_example

-> Name of the job is gromacs_example

module load gromacs/5.1.2/gcc493/impi_cuda

-> Load the environment for gromacs cuda version

cd $PBS_O_WORKDIR

-> Change to current working directory

mpirun mdrun_mpi -pin on -ntomp $OMP_NUM_THREADS -s inputfile.tpr >mdrun.log

-> Execute mpi version of gromacs command

* To submit the job use the qsub syntax in example 3

* In this command notice mpirun does not have any arguments as PBS Pro is tightly integrated with mpirun command

* In case of application enabled with openmp and MPI, it is suggested to use mpiprocs times ompthreads less than or equals to 24.

Example 6: Array Job

To maximize the usage of a serial application it is useful to submit array jobs. Array jobs are usually index driven where the index number is generated by PBS. The index number can be used as an input argument for the input file/application to spawn multiple jobs to run parallel.

Login node
$	cat submit.pbs ------------------------------------------------------------------- #!/bin/bash #PBS -q normal #PBS -l select=1:ncpus=24:0mem=96G #PBS -l walltime=16:00:00 #PBS -P Personal #PBS -N Array_Job #PBS -J 1-10 echo "Main script: index " $PBS_ARRAY_INDEX /opt/AppA –input /home/user01/runcase1/scriptlet_$PBS_ARRAY_INDEX

cat submit.pbs

-------------------------------------------------------------------

#!/bin/bash

#PBS -q normal

#PBS -l select=1:ncpus=24:0mem=96G

#PBS -l walltime=16:00:00

#PBS -P Personal

#PBS -N Array_Job

#PBS -J 1-10

echo "Main script: index " $PBS_ARRAY_INDEX

/opt/AppA –input /home/user01/runcase1/scriptlet_$PBS_ARRAY_INDEX

The jobs script is explained below

#PBS -> this is the PBS directive

#PBS –q normal

-> selecting queue as normal

#PBS –l select=1:ncpus=24:mem=96G

-> for this job 1 unit of 24 cpus and 96GB of memory

#PBS –l walltime=16:00:00

-> Requested wall time is 16 hours

#PBS –P Personal

-> Request PBS to account under Personal project

#PBS –N Array_Job

-> Specify the name of the job is Array_job

#PBS -J 1-10

-> Array index is defined from 1 to 10

echo "Main script: index " $PBS_ARRAY_INDEX

-> for debug purpose echo $PBS_ARRAY_INDEX variable

/opt/AppA –input /home/user01/runcase1/scriptlet_$PBS_ARRAY_INDEX

-> execute the application APPA with requires the input file suffix with $PBS_ARRAY_INDEX

* To submit the job use the qsub syntax in example 3

PBS Pro Job Environment variables:

While PBS jobs are executing there are few variables that are defined by PBS which can be used during the job execution. Below table provide list of few such variables.

Variable	Description
PBS_JOBID	The identifier assigned to the job
PBS_JOBNAME	The job name supplied by the user
PBS_NODEFILE	The filename containing a list of vnodes assigned to the job
PBS_O_PATH	Value of the PATH from submission environment
PBS_O_WORKDIR	The absolute path of directory where qsub is to execute
TMPDIR	The job-specific temporary directory for this job

Environment variables can used in Array jobs

Variable	Used for	Description
PBS_ARRAY_INDEX	Subjobs	The identifier assigned to the job
PBS_ARRAY_ID	Subjobs	The job name supplied by the user
PBS_JOBID	Jobs, Subjobs	The filename containing a list of vnodes assigned to the job

Querying job

Once the job is submitted to PBS, and wanted to know the status of the job, the command that need to be used is “qstat”. The qstat command shows the lists the jobs that are currently running or jobs that are in the queue for execution.

Information about the jobs that finished more than 24 hours ago is not displayed in the output of qstat command.

Syntax: qstat <-a, -n [-1], -s [-1], -f ,-x, -w, -r, -I, ..>

To view jobs that are running or in the queue use

Login node
$	qstat ------------------------------------------------------------------- Job id Name User Time Use S Queue --------------------- ---------------- ---------------- -------- - ----- 161201.wlm01 STDIN fsg5 00:00:00 R dev

qstat

-------------------------------------------------------------------

Job id Name User Time Use S Queue

--------------------- ---------------- ---------------- -------- - -----

161201.wlm01 STDIN fsg5 00:00:00 R dev

In the above example, we can see a list of jobs that are currently being executed which are submitted by me.

All other users’ jobs are hidden and so to get the list of other users’ jobs use the command “gstat”.

Another example is demonstrated below. The argument -H provides more information such as nodes, tasks, and memory requested.

Login node
$	qstat -H ---------------------------------------------------------------------------------------- Req'd Req'd Elap Job ID Username Queue Jobname SessID NDS TSK Memory Time S Time --------------- -------- -------- ---------- ------ --- --- ------ ----- - ----- 113965.wlm01 fsg5 medium Sleep 9440 8 192 512gb 24:00 F 00:00 113966.wlm01 fsg5 medium mpitest 18335 8 192 512gb 24:00 F 00:00 113999.wlm01 fsg5 medium GMX 9600 8 192 512gb 24:00 F 00:00

qstat -H

----------------------------------------------------------------------------------------

Req'd Req'd Elap

Job ID Username Queue Jobname SessID NDS TSK Memory Time S Time

--------------- -------- -------- ---------- ------ --- --- ------ ----- - -----

113965.wlm01 fsg5 medium Sleep 9440 8 192 512gb 24:00 F 00:00

113966.wlm01 fsg5 medium mpitest 18335 8 192 512gb 24:00 F 00:00

113999.wlm01 fsg5 medium GMX 9600 8 192 512gb 24:00 F 00:00

To view the jobs regardless of the job state use the below example:

Login node
$	qstat –x ---------------------------------------------------------------------------------------- Job id Name User Time Use S Queue --------------------- ---------------- ---------------- -------- - ----- 100948[].wlm01 amber89 fsg5 0 B long 104843[].wlm01 submitTCGAediti fsg5 0 B small 135905.wlm01 ipmi-reboot.pbs fsg5 00:00:00 F dev 144581.wlm01 CFS fsg5 574:49:0 F small 158001.wlm01 clean_HK-pfg052 fsg5 0 H dev 154264.wlm01 aa fsg5 0 H gpunormal 104720.wlm01 parallelSearch. fsg5 0 H medium 136070.wlm01 ipmi-reboot.pbs fsg5 0 Q dev 154047.wlm01 CY-gpu-p fsg5 0 Q gpunormal 159191.wlm01 nap_br2 fsg5 0 Q long 127876.wlm01 U7-DDES fsg5 0 Q medium 160377.wlm01 mono_SnTe fsg5 0 Q small 161031.wlm01 STDIN fsg5 00:00:33 R dev 137437.wlm01 D1LPSPep fsg5 2172:18: R gpulong 154263.wlm01 aa fsg5 1724:47: R gpunormal 161213.wlm01 STDIN fsg5 00:00:00 R largemem 127563.wlm01 InvertedFoil fsg5 2797:41: R long 112987.wlm01 QuantumEspresso fsg5 0 R medium

qstat –x

----------------------------------------------------------------------------------------

Job id Name User Time Use S Queue

--------------------- ---------------- ---------------- -------- - -----

100948[].wlm01 amber89 fsg5 0 B long

104843[].wlm01 submitTCGAediti fsg5 0 B small

135905.wlm01 ipmi-reboot.pbs fsg5 00:00:00 F dev

144581.wlm01 CFS fsg5 574:49:0 F small

158001.wlm01 clean_HK-pfg052 fsg5 0 H dev

154264.wlm01 aa fsg5 0 H gpunormal

104720.wlm01 parallelSearch. fsg5 0 H medium

136070.wlm01 ipmi-reboot.pbs fsg5 0 Q dev

154047.wlm01 CY-gpu-p fsg5 0 Q gpunormal

159191.wlm01 nap_br2 fsg5 0 Q long

127876.wlm01 U7-DDES fsg5 0 Q medium

160377.wlm01 mono_SnTe fsg5 0 Q small

161031.wlm01 STDIN fsg5 00:00:33 R dev

137437.wlm01 D1LPSPep fsg5 2172:18: R gpulong

154263.wlm01 aa fsg5 1724:47: R gpunormal

161213.wlm01 STDIN fsg5 00:00:00 R largemem

127563.wlm01 InvertedFoil fsg5 2797:41: R long

112987.wlm01 QuantumEspresso fsg5 0 R medium

The output of the gstat command will be something similar to that shown below

Login node
$	Gstat ---------------------------------------------------------------------------------------- Node States: free : 158 job-busy : 1138 job-exclusive : 0 offline : 3 ******************************************************************************* Mon Aug 8 12:58:01 SGT 2016 Req'd Req'd Elap Job ID Queue SessID NDS TSK Memory Time S Time --------------- -------- ------ --- ----- ------ ----- - ----- 100948[].wlm01 long -- 1 24 105gb 120:0 B -- 100949[].wlm01 long -- 1 24 105gb 120:0 B -- 104720.wlm01 medium -- 66 66 586gb 02:00 H -- 104721.wlm01 medium -- 66 66 586gb 02:00 H -- 104722.wlm01 medium -- 66 66 586gb 02:00 H -- 104723.wlm01 medium -- 66 66 586gb 02:00 H -- 104724.wlm01 medium -- 66 66 586gb 02:00 H -- 104725.wlm01 medium -- 66 66 586gb 02:00 H -- 104727.wlm01 medium -- 66 66 586gb 02:00 H -- 104843[].wlm01 small -- 1 24 105gb 24:00 B -- 105116.wlm01 medium -- 66 66 586gb 02:00 H -- 105117.wlm01 medium -- 66 66 586gb 02:00 H -- 105118.wlm01 medium -- 66 66 586gb 02:00 H -- 105119.wlm01 medium -- 66 66 586gb 02:00 H -- 105120.wlm01 medium -- 66 66 586gb 02:00 H -- 105121.wlm01 medium -- 66 66 586gb 02:00 H -- 105122.wlm01 medium -- 66 66 586gb 02:00 H -- 105123.wlm01 medium -- 66 66 586gb 02:00 H -- 105124.wlm01 medium -- 66 66 586gb 02:00 H --

Gstat

----------------------------------------------------------------------------------------

Node States:

free : 158

job-busy : 1138

job-exclusive : 0

offline : 3

*******************************************************************************

Mon Aug 8 12:58:01 SGT 2016

Req'd Req'd Elap

Job ID Queue SessID NDS TSK Memory Time S Time

--------------- -------- ------ --- ----- ------ ----- - -----

100948[].wlm01 long -- 1 24 105gb 120:0 B --

100949[].wlm01 long -- 1 24 105gb 120:0 B --

104720.wlm01 medium -- 66 66 586gb 02:00 H --

104721.wlm01 medium -- 66 66 586gb 02:00 H --

104722.wlm01 medium -- 66 66 586gb 02:00 H --

104723.wlm01 medium -- 66 66 586gb 02:00 H --

104724.wlm01 medium -- 66 66 586gb 02:00 H --

104725.wlm01 medium -- 66 66 586gb 02:00 H --

104727.wlm01 medium -- 66 66 586gb 02:00 H --

104843[].wlm01 small -- 1 24 105gb 24:00 B --

105116.wlm01 medium -- 66 66 586gb 02:00 H --

105117.wlm01 medium -- 66 66 586gb 02:00 H --

105118.wlm01 medium -- 66 66 586gb 02:00 H --

105119.wlm01 medium -- 66 66 586gb 02:00 H --

105120.wlm01 medium -- 66 66 586gb 02:00 H --

105121.wlm01 medium -- 66 66 586gb 02:00 H --

105122.wlm01 medium -- 66 66 586gb 02:00 H --

105123.wlm01 medium -- 66 66 586gb 02:00 H --

105124.wlm01 medium -- 66 66 586gb 02:00 H --

Job Dependencies

Users can specify dependencies between their jobs, such as:

Specify order of execution
Execute the next job only if previous job finished
Place jobs on hold until a particular job starts or finishes

Syntax: qsub -W depend=<type>:<arg_lists> <job ID> myjobscript.sh

For multiple dependencies, <arg_list> items are colon-separated

Example: qsub -W depend=afterok:1.pbsworks:2.pbsworks myjobscript.pbs

Deleting jobs

When the job is in queued state or running state, it may be necessary to terminate the job. In case if the job need to be terminated, the command to be used is qdel.

Syntax: qdel <job id>

Usage: qdel 2.wlm01

To delete the job forcefully, -W argument need to be supplied to the qdel command.

Syntax: qdel -W force <job id>

Usage: qdel -W force 2.wlm01

Project summary

To check the allocation of CPU hours in the system, the command to use is project_summary.

Syntax: project_summary <project id>

Usage: project_summary 11000000

When project ID is not specified, the summary displayed is only personal project.

NSCC-PBSPro-QuickstartGuide

Executive Summary

Intended Audience

Assumptions

Document Conventions

Support

Introduction​​ to NSCC Job Scheduler (PBS Pro)

Robust Workload Management

Advanced Scheduling Algorithms

Reliability, Availability and Scalability

Getting Started with PBS Pro

PBS Queues

External Queues:

Internal Queues:

System wide restrictions

Job Submission

PBS Pro Job Environment variables:

Querying job​​

Job Dependencies

Deleting jobs

Project summary

Introduction to NSCC Job Scheduler (PBS Pro)

Querying job