コンテンツ

  1. HOME
  2. FAQ
  3. Computing Services
  4. Use of a supercomputer

Use of a supercomputer FAQ

Users may log in to the Systems using SSH (#22 port of TCP protocol) with no access restriction. Please contact your network administrator. Note that password authentication is prohibited and SSH authorization is required to log in to the System.

For how to log in to the system, see this page.

SSH key authentication is required to log in to the systems Thus, users are required to generate a public/private key pair and register the public key via the User Portal.

Please contact the Supercomputing Services of ICT Support Center. When contacting us by email, please describe your user ID and name using the email address you registered with us.

Once successfully logged in to the Systems, registering public keys is disabled on the User portal To add a separate public key, directly add its key in ~/.ssh/authorized_keys after logging to the Systems.

This error message appears when an unauthorized user attempted to log in the system other than the one which such user completed the service contract.

Private keys are no longer available without passphrases. Please re-generate a public/private key pair. Note that once successfully logged in to the System, registering public keys is disabled on the User portal. Please contact the Supercomputing Services of ICT Support Center. When you contact us by email, please describe your user ID and name using the email address you registered with us.

Exceed onDemand does not support key authentication. Thus, Port Forward is required to use along with Exceed onDemand. For details, see Logging In With Exceed onDemand.

For existing users who completed the application process for continued use, the password remains the same as the one which you were using in the previous year, not the initial password printed in the Certificate of Approval for Supercomputer Use.

The error message means the host name you entered is incorrect. Please make sure to enter the "localhost" in the Exceed on Demand start-up window. See the Logging In With Exceed onDemand for details.

Please note our services may be temporarily suspended during system failure or scheduled maintenance. During such periods, users are unable to log in to the Systems. We notify users of the service suspension via the Email Newsletter and on our website in a timely manner.

You may log in to the System until the end date of the fiscal year (before the year-end maintenance scheduled in late March). It is recommended users back up all necessary data including resulting computation data created immediately before its date. After its date, users may not login to the Systems but are allowed to transfer files during a grace period until the end of May for the following fiscal year.

Q: I ended the use at the end of the previous fiscal year, but I want to extract some data containing a large amount of data during the grace period until the end of May when transferring the data is allowed. Is it possible to log in to the System so I can compress the large data files?

As a rule, we ask users to back up all necessary data by the end of the fiscal year, but if it is necessary to log in to the System due to circumstances, please contact us via the Form.

Q: I stopped using the supercomputer services at the previous fiscal year end, and I will download the remaining data during a period of grace for transferring the data (until the end of May). But I forgot my private key for logging in to the supercomputer, so can I re-register a new public key?

A: We provide a contact form, the form authorizes any expired user ID. Please include your expired user ID, password, and public key that you require to re-register, in the "お問い合わせ内容" field.

Please configure the SSH client and set the environment variable to LANG as below:

Setting up an SSH client (Windows)
    • In PuTTY(Japanese)

Select ウィンドウ > 変換 > 文字コードの設定, and then select UTF-8.

    • In TeraTerm

In the menu, select 設定 > 端末 > 漢字-受信 and 漢字-送信, and then select UTF-8.

Setting up the environment variable, LANG.
In the case of tcsh
 $ setenv LANG "ja_JP.UTF-8" 
In the case of bash
$ export LANG="ja_JP.UTF-8"

At log-in, you can automatically configure the software by describing:

(When using tcsh) ~/.tcshrc

(When using bash) ~/.bashrc

Or when configuring the Gaussian, load the "module load gaussian09".

You may not change the limit because we set the quota limit for the home directory. In the Personal, Group, or Private Cluster course, the large volume disk (LARGE) is also available as well as the home directory.

The elapsed time for queue of group and personal course is set to one hour by default. To change the upper limit of the elapsed time, please use the -W option.

But the elapsed time for the entry course queue(eb) is limited to one hour, and it cannot be exceeded.

In the Group Course, queues are shared among the users of each group. When the amount of computational resources for the job exceeds the minimum amount of allocated cores for the queue, each user ensures resources at a time. Accordingly, it sometimes takes a long time to start to run a job depending on usage status of other users of the whole system.

During a large-scale job is scheduled to run, other jobs may be placed in a pending status depending on the scale of the large-scale job. See the Large-Scale Job Schedule.

The priority queue is used only for helping jobs at the time of failure. Thus, you fail to submit any job directly to the priority queue.

This error message appears when the requested amount of computation resources exceeded the upper limit of the resources allocated per job for each queue.

The upper limit per job for each queue means the number of nodes notified via E-mail. In the Group Course, users also may check it as shown in STD using the qgroup command (System A, System B and C).

When requesting resources exceeding the memory capacity per core (see below), the amount of cores corresponding to the memory capacity is assigned.. For example, when executing a program with 4 processes that request 3,000 MB memory per process in the System A, 3 cores divided 3,000 MB by 1,355 MB with any fraction rounded up, the resources of 12 cores (3 cores x 4 processes) is assigned.

SystemMemory Capacity Per Core
A 1,920MB
B 3,413MB
C 42,666MB
E 3,072MB

This error may occur when the job script has Windows newline (CR+LF). Please convert a newline from Windows to UNIX by using the dos2unix command before submitting the job.

If you type qdel command but an error message is shown and jobs are not deleted, please try "-W force" option.

Syntax: qdel -W force {job number}

Please check whether the following error message is in the standard error output file.

/var/spool/PBS/mom_priv/jobs/1105396.jb.SC: line 28: 179454 Killed 

If the above message is seen, the amount of memory, which is specified by "#QSUB -A" option in the jobscript, is not enough to complete the job.

In the maintenance taken on Oct. 2017, a bug, which the compute nodes cannot correctly restrict the amount of memory specified by the jobscript, was fixed.

After the maintenance, a job which uses the amount of memory exceeding the restriction by #QSUB -A will be aborted with the error message.

If your job is involved in the above error, please adjust the description of job relevant to the memory and re-submit the job.

On KUDPC, due to the prevention of disturbing other users' jobs, the total capacity of standard output files of jobs running at the same time is restricted to 8GB.
If exceeding 8GB, the jobs running at the same time are forced to be halted.

If your program will output a lot of data, please make the program redirect output to files as follows.

Ex) on System A:

aprun (*snip*) ./a.out > output.log

Before using the application, you are required to set the environment using the module command. For details on how to use, see this page.

SAS is available in only a single node of whole login nodes in the system B and C. When you use it, please access to sas.kudpc.kyoto-u.ac.jp.

If maple doesn't work in the GUI mode, it sometimes works normally by exceeding the memory size(virtual memory) of the process limit before startup.
For details of process limit, see this page.

If this error occurs, tecplot sometimes works normally by using -mesa option when startup.

The -mesa option emulates the features of OpenGL graphics library through the software.
Thus, it decreases the rendering speed a little.

And when you got the error message "Failed to create OpenGL Context," tecplot sometimes works normally by using -mesa option when startup.

Application software often start in a GUI mode normally, by modifying the configuration file(Xconfig file) that is specified when logging in to the system with the Exceed onDemand.

Before logging in to the system, enter your user ID and password, then specify the Passive_noglx.cfg for the Xconfig file in the connection information.

When using the application software, please use the xrun command with the GUI, or use the tssrun command without the GUI.

Note that the default walltime (max execution time) is 1 hour. If you want to expand the walltime, please use -W option.

For details on the commands, see "Interactive Processing".

The compile commands are different between the system A and the others. In the system A, please use the ftn command instead of the ifort.

relocation truncated to fit: R_X86_64_32S against `.bss'
additional relocation overflows omitted from the output

This message indicates that the memory that is statically allocated in the program may exceed 2GB. Please specify each one of the following options when compiling.

System A:
[Cray compiler] -h pic -dynamic
[Intelcompiler] -mcmodel=medium -shared-intel -fpic -dynamic
[PGI compiler] -mcmodel=medium -dynamic
[GNU compiler] -mcmodel=medium -fpic -dynamic

System B:
[Intel compiler] -mcmodel=medium -shared-intel
[PGI compiler] -mcmodel=medium
[GNU compiler] -mcmodel=medium

ftn-73 crayftn: ERROR MAIN, File = sample.f90, Line = 6, Column = 30 This control item specifier is not allowed for this I/O statement.

A conversion specifier of OPEN statement does not exist in FORTRAN standards, and is not supported by the Cray compiler. Thus, both file formats(big-endian and little-endian) cannot be loaded in the same program by the Cray compiler.

lib-4211 : UNRECOVERABLE library error A WRITE operation tried to write a record that was too long.

Encountered during a sequential formatted WRITE to unit 11 Fortran unit 11 is connected to a sequential formatted text file: "fort.11" Current format: 1000 FORMAT(1000(f7.5,1x))

Cray Fortran compiler sets an output restriction to 1024 characters per line by default. To avoid this error, please specify the number of characters per line by using the RECL specifier in the OPEN statement. (e.g., recl=2048)

sample.s:837: Error: 0x9ab5c0f0 out range of signed 32bit displacement

This message indicates that the memory usage of local variable within the FORTRAN subroutine may exceed 2GB. Specify the -ev option when compiling.

MPI_Irecv(110): Null Datatype pointer

This message indicates that MPI program include "MPI_LONG_DOUBLE".Cray compiler does not support "MPI_LONG_DOUBLE", please use "MPI_DOUBLE". When using MPI in the system A, Intel compiler (PrgEnv-intel) and GUN compiler (PrgEnv-gnu) support "MPI_LONG_DOUBLE".

Any supercomputer user can use an application software without attending a programming seminar. Application software that we offer are different depending on the system. For details, see "Application Software".

Seminar material can be downloaded at the programming seminar list after the seminar. The material that cannot be provided as an electronic file will be offered as a paper document for system users who need it, if there will be some left over. If you would like to get the seminar material of a paper, please contact the Inter-University Services Section.

You can find the information about your session, TSS, and batch through the user portal. The payment managers can view the user's information, and the group managers can view the batch information of their group members.

Please set MALLOC_ARENA_MAX as 1.

Example: in bash environment

export MALLOC_ARENA_MAX=1

 

Copyright © Institute for Information Management and Communication, Kyoto University, all rights reserved.