Frequently Asked Questions

Questions about Univa Grid Engine

General Questions

Questions about Univa Grid Engine

An error is output during qstat command execution in Grid Engine.

% qstat -f
error: commlib error: got select error (Connection refused)
error: unable to send message to qmaster using port 6444 on host "c722i": got send error

This is because the Grid Engine master daemon is being switched off for preventive maintenance or there is a high load on the computer, or the master daemon is not working due to a problem. Switching off the master daemon is executed automatically to address stable operation.
Please wait for a while until the GE master daemon switching off is completed. Switching off is normally completed in about 5 minutes, and you should be able to issue the qstat command again. If the error occurs due to a stoppage caused by a problem, the restoration status is updated as appropriate on the following page. Please check this Web page.

Can I input jobs into an exclusive queue?

Users who have applied for exclusive cores can use exclusive queues. By inputting the job without specification of the option during qsub, it will be input to mjobs.q or an exclusive queue depending upon the queue usage status. To input only into an exclusive queue, the job needs to be input by specifying the option as follows: % qsub -q [name of exclusive queue] [name of shell script] Exclusive queues do not have an upper limit for the execution period. By specifying as shown above when ljobs.q,lljobs.q is congested, it is possible to execute jobs with long execution periods using exclusive queues.

Execution of jobs queued for nighttime/holidays

Jobs can be input into the nighttime queue and holiday queue even when it is not the corresponding period for the nighttime queue or the holiday queue. Input jobs will be executed during the nighttime queue or holiday queue periods. Jobs that are in execution at the end of the period will be halted, and rescheduled for re-execution. Please note that data will be added to the files recording the standard output and standard error output for the job when the job is re-executed.

For details regarding the use of nighttime and holiday queues, please see here.

General Questions

Can I log into the supercomputer system without a password?

Password authentication is required by default when ssh is used.
By executing the following, you can set up key-exchange authentication. As the home directory is shared within the supercomputer system, a password is not be required when logging onto all hosts within the supercomputer system by executing the following in the same host.
* Creating the public key and the private key
% ssh-keygen
Generating public/private rsa key pair.
Enter file in which to save the key (/home/XXXX/.ssh/id_rsa):
Enter passphrase (empty for no passphrase):
# Press the enter key without inputting the pass phrase.
Enter same passphrase again:
#Press the enter key without inputting the pass phrase in a similar fashion.
Your identification has been saved in /home/XXXX/.ssh/id_rsa.
Your public key has been saved in /home/XXXX/.ssh/
The key fingerprint is:
*Registering the public key
% cat .ssh/ > .ssh/authorized_keys

I could not attend the lecture meeting. Can I get the materials later?

The materials used at the lecture meetings are posted on this page for public access. Videos of the lecture meetings are also available. (A supercomputer account is needed to view the materials and videos.)

I have forgotten my password

You can get a new password by applying for a change of password by sending the supercomputer support section an e-mail informing us of the situation. Support personnel will contact you to confirm matters.

When we contact you, fill out the user section and the research supervisor section entering password replacement as the reason for application in change as the type of application on the new user registration form, and send it to the supercomputer support by fax. Please note that it takes three to five days from application to your receipt of the password, as it will be sent to you by mail.

I cannot log onto the supercomputer

You may not be able to log into the system if there is a problem with the supercomputer. For information about such problems, please check here.

Error when running Java

On July 23rd, 2013 we added the JAVA_TOOL_OPTIONS default environment variables for running Java. Due to this, errors do not occur now when running simple Java programs. $ echo $JAVA_TOOL_OPTIONS
-XX:+UseSerialGC -Xmx64m -Xms32m

When running Java, if you have not specified the maximum heap size and the initial heap size to be used, then by default the system will try to reserve 1/4 of the heap size of its physical memory. On nc, this is 8 GB, on ncl this is 36 GB; therefore, an error occurs when its usage goes over the shell memory limit (by default Shirokane2: 1 GB).

Please make sure you specify the heap size in line with your Java application (e.g., ${JAVA} -Xms2g -Xmx2g ), and change the shell script memory limit when using qlogin or qsub (e.g., qlogin -l s_vmem=3G -l mem_req=3). Java's -XX:+UseSerialGC option is also effective. Furthermore, if you set the heap size to be the same as the memory limit, the whole job may still go over this limit, so please make sure that the memory limit is greater than the heap size.

Top of Page Top of Page

The University of Tokyo The Institute of Medical Science

Copyright©2005-2015 Human Genome Center