Interactive Processing

Interactive processing is performed with an LSF and PBS interactive batch system. The tssrun command and the xrun command are provided to easily run a program in an interactive mode.
You can use these commands by entering a program and an argument that you want to run after these commands.

Command Purpose Example Notes
tssrun Running a program on a computation node. tssrun ./a.out
xrun Running a GUI program on a computation node. xrun ./a.out Only available in the Exceed onDemand or other X server software.
  • By entering the command, some messages are displayed, then the program is started and a result is displayed.
  • When a computation node has too much processing, the program may not be started soon after messages are displayed.
  • You can also use an interactive mode in a group course queue. For details, see Use Interactive Mode in Group Course Queue.

Example 1: Running a MPI program in the System A

qsub: waiting for job 1249.sdb to start
qsub: job 1249.sdb ready

%%% Aprun Wrapper
%  launched with the following command
%
%   /opt/cray/alps/default/bin/aprun -n 4 -d 1 -N 1 -j 1 -cc 0 ./a.out
%
%%% end of Aprun Wrapper message

    myrank = 0: Hello, World!
    myrank = 1: Hello, World!
    myrank = 2: Hello, World!
    myrank = 3: Hello, World!

Application 380396 resources: utime ~0s, stime ~0s, Rss ~4624, inblocks ~0, outblocks ~0

qsub: job 1249.sdb completed

Example 2: Running a MPI program in the System B

qsub: waiting for job 1249.sdb to start
qsub: job 1249.sdb ready

%%% Aprun Wrapper
%  launched with the following command
%
%   /opt/cray/alps/default/bin/aprun -n 4 -d 1 -N 1 -j 1 -cc 0 ./a.out
%
%%% end of Aprun Wrapper message

    myrank = 0: Hello, World!
    myrank = 1: Hello, World!
    myrank = 2: Hello, World!
    myrank = 3: Hello, World!

Application 380396 resources: utime ~0s, stime ~0s, Rss ~4624, inblocks ~0, outblocks ~0

qsub: job 1249.sdb completed

Example 3: Using an interactive mode in a group course queue

qsub: waiting for job 1249.sdb to start
qsub: job 1249.sdb ready

%%% Aprun Wrapper
%  launched with the following command
%
%   /opt/cray/alps/default/bin/aprun -n 1 -d 1 -N 1 -j 1 -cc 0 ./a.out
%
%%% end of Aprun Wrapper message

    myrank = 0: Hello, World!

Application 380396 resources: utime ~0s, stime ~0s, Rss ~4624, inblocks ~0, outblocks ~0

qsub: job 1249.sdb completed

Example 4: Running a GUI program

$ xrun matlab
Job <80279> is submitted to queue <tb>.
<<Waiting for dispatch ...>>
<<Starting on gb-0003>>
(X application is started.)

When performing a process parallelism or a thread parallelism, you can specify a parallel number using the command option. You can also specify the limit of memory use, elapsed time limit for a program, and CPU time limit. For details, see Limitations on Interactive Mode.

Option Purpose Example
-A p=X :t=X:c=X:m=X Specifying resources for allocating to a job. -A p=4:t=8:c=8:m=2G
-W HOUR:MINUTE Specifying the upper limit of the elapsed time. (Hour:Minute) -W 1:0
  • Specifying resources

You can specify computing resources for allocating to a job using the -A option. There are four variables as arguments of the -A option: p, t, c, and m. These are concatenated with a colon(:). The following table shows the purposes of these four variables.

Argument of the -A Purpose
p=X Allocated process when executing jobs
t=X Allocated thread per process when executing jobs
c=X Allocated CPU cores per process when executing jobs
m=X M /X G /X T Upper limit of allocated memory per process when executing jobs(unit: M, G, T)

When the -A option is disabled, the values of p, t, and c set to 1, and the value of m is set to the number of memory on node divided by core on node. If even one of these four variables is not specified, the default value is set. The value specified at t is automatically set to the environment variable “OMP_NUM_THREADS.”

Supercomputer users who use Service Courses except for the Entry Course can run a program in an interactive mode using a batch processing node allocated at their service courses. In this case, the program is not limited to the following section “Limitations on Interactive Mode,” and the available resources and the elapsed time limit are specified by the queue you specified.

Option Purpose Example
-q QUEUENAME Specifying a batch processing queue. -q gr19999b
-ug GROUPNAME Specifying a group. (For Group course) -ug gr19999

The following limitations apply when you run a program in an interactive mode (except you run in a personal course queue or group course queue)

  • Node sharing

Nodes for interactive processing are provided to each System (the System A, B, C), all users of each System share each node. The program is run exclusively with the occupied CPU cores, so it may take time to be started when the node has too much processing.

  • the number of jobs at the same time

One user can RUN one job at the same time. Other jobs are forced to be PEND.

  • CPU core limit

CPU cores that can be available at the same time have an upper limit number for a single user.

System A B C
CPU Core Limit 68 36 72
  • Memory limit

The amount of memory used by the program is limited to about 2 GB in the System A, about 4 GB in the System B, about 45 GB in the System C, and about 3 GB in the System E. The -A option allows you to exceed the amount of memory up to the upper limit.

System A B C
Standard Memory 1355M 3413MB 42666MB
Maximum Memory 90G 120GB 3000GB
  • Elapsed time-limit of program

A program that is run in an interactive mode is terminated forcibly when its elapsed time exceeds the limit time of standard. The -W option allows you to exceed the time up to the upper limit.

System A B C
Standard Elapse Time (hour:minute) 1:00 1:00 1:00
Maximum Elapsed Time (hour:minute) 24:00 24:00 24:00

Job information can be checked after interactive execution by the tssrun or xrun command.

  • System A/B/C
Command Option Purpose Example of Execution
qstat -xf Checking the information of jobs executed in an interactive mode. qstat -xf 1405.sdb

Job ID needs to be included after the command and option. Job ID is indicated as "xxx.sdb" in the message that is displayed when executing the tssrun or xrun command.

After 24 hours from the time the interactive execution was completed, the job information cannot be checked.

If the interactive execution is abnormally terminated, the reason of the abnormal termination will be shown in the job information. The examples in the following are on the System E.

  • The message shown when terminated by the memory limit.

    [NID 00181] 2016-10-07 16:04:40 Apid 391976: OOM killer terminated this process.
  • The message shown when terminated by the elapsed time limit.

    PBS: job killed: walltime 3602 exceeded limit 3600


Copyright © Academic Center for Computing and Media Studies, Kyoto University, All Rights Reserved.