Frequent Asked Questions (FAQ)
General
How could I know the computing environment (including the library, compiler, etc.) of DiCOS ?
If you are using slurm_ui, the software stacks are maintained with the environment-modules software. Please follow this link for the instruction.
Please contact us by e-mail (DiCOS-Support@twgrid.org) or fill the form in this link for any required system software.
How to use the Command Line Interface in DiCOS?
Login to dicos-ui05.grid.sinica.edu.tw by using the account/password used in the DiCOS Web User Interface.
When I login in dicos-ui04.grid.sinica.edu.tw, I got port: 22 connection denial. Why?
It’s most likely that you are banned due to multiple login failures with your IP address. Please wait for 2 to 3 hours then try again. Or use other machines to login. If your forget your password, please use this link to reset your password.
How to reset my DiCOS password if I forget it?
If your forget your password, please use this link to reset your password. A follow-up mail will be sent to your registered e-mail address, please follow that up to reset the password.
DiCOS Account
I could login in dicos-ui04.grid.sinica.edu.tw with my account yesterday, but I could not login in today with the same credential. Why?
If you are not banned by multiple wrong credential inputs in a short time (see above), one possibility is your DiCOS account is expired. The default lifetime of each DiCOS account is 1 year, you will be notified by e-mail from DiCOS system prior to the expiration.
However, sometimes, we have some issues that the e-mail from ASGC to gmail will sometimes not delivered, so if you are using gmail, you will possibly miss the account extension mail. Please go to: https://canew.twgrid.org/ApplyAccount/nocertModify.php, and input your original registered e-mail. Then reset your password with the instructions on the e-mail to re-activate your account.
We will try to solve the e-mail delivery issue soon to fix this issue. We are sorry for the inconvenience at this moment.
DiCOS Job System
When I use dicos ddm download command, it failed with the following messages: [2022-04-12 03:35:31,188 ERROR] Failed to download. What happened?
There should be something wrong with our DDM subsystem. You can use dicos ddm ping to check if it returns: Ping Rucio OK ! message. If not, please contact to DiCOS-Support@twgrid.org to report the issue. Thank you.
How can I debug the jobs submitted by dicos job submit? Could I have more information on my jobs?
For example, you could get the PandaID (Job ID) by dicos job status. However, it only shows the status of the jobs. To see the detail information of your job, you could use:
dicos job status -p <PandaID>
and, a JSON string will be returned with detail information.
Besides, you could try to forcely get the outputs of your job by:
dicos job getoutput -f <PandaID>
and see if there is any hints. Or use:
dicos job getoutput -f --log <PandaID>
to download the outputs together with the logs, and try to debug yourself. If you still could not find what’s going on with your jobs. You could contact to DiCOS-Support@twgrid.org for consultation.
Slurm
How do I know the current usage/vacancy of specific machine of slurm?
For example, we could use the following command to get the status of hp-teslav01:
sinfo --format="%10n %.14C %.10T" -n hp-teslav01
How could I use python3? What should I do if I would like to install some python packages from pip?
In our user interface (UI) and worker nodes, we have CVMFS file system mounted as the software repository. You could initialize anaconda3 (python3) with the following command before you start using python3 (in the UIs and in the worker nodes):
source /cvmfs/cvmfs.grid.sinica.edu.tw/twgrid/anaconda3/setup.sh
Then you will have anaconda3 python3 available (python 3.8.5). You could include this line in the very beginning of the script. The numpy package is by default included in the anaconda3 bundle.
If you are going to install some customized packages, we recommend you to installed it in your home directory. For example:
pip3 install bilby --user
pip3 install gwpy --user
Then you will have these packages available in your home directory. But if you are using dicos submit, then we will need to install the missing packages for you. Please contact to DiCOS-Support@twgrid.org for help.
I encounter an error when I compile my CUDA code on slurm-ui01. The error message is cannot open source file “cuda_runtime_api.h”. Could you help me solve this problem?
The login node (slurm-ui01) doesn’t have GPU resources, and hence don’t have cuda library installed. So you could not build your cuda apps in slurm-ui01. We recommend you to use the DiCOSApp that have GPU resource support, for example: Jupyter Lab GPU 1080ti, to build your cuda apps.
Please use the terminal in jupyterlab and check CUDA_PATH environment variable is set to: /usr/local/cuda to build your applications.
How to use nvcc in slurm? Could you give me some instructions?
There are V100 and A100 GPU resources available in slurm QDR4. If you are going to use nvcc, after login in slurm-ui01, please load the corresponding module as follows:
module load nvhpc_sdk/20.11
Then nvcc and cuda-gdb, and pgcc (openacc compatible compiler) will be available for your usage. And if you are going to run your cuda/openacc enabled code, you need to submit your job to “v100” or “a100” partitions of slurm, e.g.
srun -p a100 --gres=gpu:2 ./my_cuda_executable my_arg1 my_arg2
DiCOSApp
How to get a CryoSPARC™ license?
CryoSPARC™ and cryoSPARC Live™ are available free of charge for non-profit academic use. Please go to https://cryosparc.com/download to apply your own CryoSPARC™ license and then import it when you launch CryoSPARC™ application.
Recently, I started a CryoSPARC™ DiCOSApp instance, but I forget the account and password setup by me when I started CryoSPARC™ App. How should I do?
CryoSPARC™ will store your login information in the CryoSPARCV3DB directory in your home directory (/dicos_ui_home/<your_account>/CryoSPARCV3DB). You could try to delete your opened CryoSPARC™ instance, and then delete the CryoSPARCV3DB directory, e.g.
rm -fr /dicos_ui_home/<your_account>/CryoSPARCV3DB
and then restart a CryoSPARC™ instance again.
How to do plugins installation in JupyterLab of DiCOSApp?
Open Terminal in JupyterLab
For example (run command in Terminal):
(python2.7)$ pip install --user {opencv-python(plugin)} (python3.x)$ pip3 install --user {plugin}
Plugins will be installed under:
/dicos_ui_home/{username}/.local/lib/python2.7/ /dicos_ui_home/{username}/.local/lib/python3.x/
and you can start using the JuyterLab Plugins