Welcome to the FAQ and answers !
What is this?
- Here we will keep track of the most Frequently Asked Questions and answers.
Introduction.
How do I start working on Shark the right way?
- Read the intro.
Why is the Shark cluster called Shark?
- Read the history of Shark.
Howto, tips and tricks.
Are there any commands to summarize the cluster's usage/availability ?
- Try to run unique-user-slots.sh or slotsfree.pl script.
Can I install any program?
- No, not yourself. However, you can request everything with the administrator using the ticket system.
Are there reference genomes available?
- Yes, you can find them with descriptive readmes in /usr/local/Genomes/.
What was that one command for that one thing again?
- Read qsub commands.
Can shark change directory automatically when I log in?
- Yes. You can add commands to the bottom of the .bashrc file in your home dir. These will be executed whenever you log into shark.
Can shark notify me when my jobs are done?
- Yes, either by email or with the notification system on your local machine.
Is R installed on the new Shark?
- Yes, You need to load the module first before you can use R. List all available module
module avail
The load the R module version you want to use.
module load R/3.1.1
Howto set your own R library location
Create a writable R directory in your /exports directory.
set the environment variable R_LIBS=/to/your/created/R/path
Now you can use this path for your own library's.
To permenently set this variable edit your ~/.bashrc file and add
export R_LIBS=/to/your/created/R/path
Can I restrict java to a fixed number of cpu's?
- Yes, java -XX:ParallelGCThreads=1 (set 1 to the number of cpu you want)
- To restrict the memory java uses : java -Xmx2048m -Xms2048m
When stuff doesn't work.
why won't my job start immediately?
- The cluster uses a ticketing system for the prioritisation of jobs. Tickets are distributed over the departments by ratio of investment. Any overcapacity is available for incidental processing or for people that have used up their tickets. If you find that your jobs are not scheduled in a timely manner, your department has probably used its share of tickets. If this happens on a structural basis, investment from your departement is the solution.
- show a formated sharetree
qconf -sst
Why won't my job run?
- Read error queues.
Why can't my qlogin be scheduled?
- Whenever this happens it's most likely that there are no slots available in your queue(s). Run qstat -u "*" or unique-user-slots.sh and view the usage of the shark cluster. You can still submit your job with qsub, the fair-share policy will take care that your jobs gets scheduled according to your rights.
Why is my job killed when I my script uses more then 4G of memory
- Shark has a default h_vmem=4G value set. If your job exceeds the memory threshold of 4G Open Grid Scheduler will auto kill that job. You should specify a higher h_vmem limit if your jobs needs more then 4G of memory like this
qsub -l h_vmem=xxG
The G means Gigabyte.
What does this exit code means?
- Read the manual for accounting go to exit_status.
man accounting
exit_status
Exit status of the job script (or xxQS_NAMExx specific status in case of certain error conditions).
The exit status is determined by following the normal shell conventions.
If the command terminates normally the value of the command is its exit status.
However, in the case that the command exits abnormally, a value of 0200 (octal), 128 (decimal) is added
to the value of the command to make up the exit status.
For example: If a job dies through signal 9 (SIGKILL) then the exit status becomes 128 + 9 = 137.
Why does my script not run or returns an error /bin/bash^M: bad interpreter: No such file or directory
- your script has been edited with an editor under windows, your line breaks for windows are
^M
characters.
You can view your script file and check if you have these characters with the command:
cat -v <script_name>
#!/bin/bash^M
#$ -S /bin/bash^M
#$ -q all.q^M
#$ -cwd^M
#$ -V^M
#$ -j Y^M
^M
^M
date^M
hostname^M
to remove the ^M
use the command dos2unix <file_name>
- If you submit a script with the
^M
with qsub your job will get in an error stateEqw
If this happens you can see with the commandqstat -j <job_ID>
and error at the line : error reason
execvp(/var/spool/sge/silvertipshark/job_scripts/<job_id>, "/var/spool/sge/silvertipshark/job_scripts/<job_id>") failed: No such file or directory
Is there a backup for the /exports directory
- No the isilon storage has no backup, there is no way to retrieve the files, Its is your own responsibility to backup your files on Shark.
I can not log into shark with my RESEARCHLUMC\username
In the email send to you the username has RESEARCHLUMC\username as an username construction.
This only works if you login into a windows workstation attached to the RESEARCHLUMC domain
or if you change your password on the change password site https://passwd.researchlumc.nl/
For all shark related usernames DO NOT use the RESEARCHLUMC\username construction leave the RESEARCHLUMC\ out,
hence your username for shark and the SSH jumpserver is : username
Can I run on specific nodes?
Yes you can give a list of hosts where you would like to run on with:
-l hostname="host1|host2|host3|host4"