To get access to the DCE, you need an account. There are two types of accounts : accounts for projects and accounts for lab works. In any case, it is the responsibility of your teacher to request for the account creation, we will not accept account creation requests from students directly.
As a teacher:
for requesting accounts for a labwork, you have to fill the YAML file tp.yml
for requesting accounts for a project, you have to fill the YAML file project.yml
Once filled, you must send your form to dce_support(arobase)groupes.renater.fr
Using a SSH client on Linux/MacOS¶
You can connect to the DCE, using SSH with the following command :
user@mycomputer:~$ ssh firstname.lastname@example.org Password: [username@chome ~]$
When the SSH tunnel is established, you are connected to the gateway of the DCE front-end where you can ask for an interactive session or launch a batch job.
You can ask for an interactive session by requesting a node on the interactive partitions. For example, to start bash on GPU node for an interactive session of 30 minutes:
[username@chome ~]$ srun -p gpu_inter -t 00:30:00 --pty bash [username@cam01 ~]$
You can also submit long running jobs in batch by calling
sbatch following by your job description.
[username@chome ~]$ sbatch mybatch.job Submitted batch job 500 [username@chome ~]$ squeue -u username JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON) 500 gpu_prod_night linear username R 0:15 1 cam15
Using a SSH client on Windows (PuTTY)¶
To access the DCE from a Windows machine with a SSH client, you can use PuTTY.
When asked to configure your connection, you have to fill the "Host Name (or IP address)" field with "chome.metz.supelec.fr" and ensure that the "connection type" is set to "SSH".
When Putty is configured correctly, you just have to click the "Open" button and a refer to the previous section in order to log in.
Using Visual studio code¶
You can access interact with the compute nodes in visual studio code using the remote ssh extension. It requires you to install visual studio code locally on your machine, the remote ssh extension from the studio code marketplace and to configure it.
This configuration will allow you to edit your code stored on the DCE machines and to execute it within a terminal running on a compute host. A video showing this in action in available in the examples section.
In addition to editing and running your code, there are several extensions you might be interested in :
- the notebook extension
- the tensorboard extension
Working with dcejs (Linux/MacOS/Windows)¶
We provide a multi-platform tool for connecting to the DCE called dcejs.
You can download the latest release of dcejs from the release page and install it on your system. For using graphical interface you additionally need to install some packages on your system
First connection with dcejs¶
The first time you connect to the cluster, dcejs will generate a SSH key which will be used to latter authenticate yourself. When generating the ssh key, dcejs will prompt you for a sshkey password you are free to decide but must remember. Then it will prompt you for your password to connect to the DCE for posting the key on our servers.
The next time you will connect to the DCE, dcejs will only ask you for the ssh key password.
Making a new allocation¶
For making a new allocation, you need to click on the "+ New" button. There you can allocate a node either with a reservation code (which is given to you by the teacher during a planed labwork session) or without a reservation code (which is the case when you finish you labwork in your free time or for your projects).
If you have a reservation code, you need to provide it and you are done. If you do not have a reservation code, you need to select a partition, specify an allocation time which must not exceed the maximal allocation time for the selected partition and you are done.
Users, if they want, may also customize the allocation by specifying advanced settings.
Starting a graphical interface on the allocated remote¶
We provide two softwares for having a graphical interface on the booked node : VNC and NoMachine. Both of them require you to install, on your computer, a client. For VNC, we advise you to install the TigerVNC client (there are other implementations of VNC but as we have installed the tigervnc server, you will benefit a better experience by running the tiger vnc client). For NoMachine, you do not have the choice and must install the nomachine client.
Once your computer is installed, you can start either VNC or NoMachine by cliking on the "Actions" button along your allocated job and then select the appropriate action.
After some times, you should see displayed on your allocation line, either below VNC or NoMachine an URL like "localhost:5901" or "localhost:4000". That URL might differ so carefully read the interface. The displayed URL is the one to be used by the viewer client.
Ending an allocation¶
Once finished, you can stop your allocation by clicking on the "Actions" menu along your allocation. This is a good practice to free the node so that other users can make use of it. In any case, the scheduler will stop your allocation after the walltime you specified at the allocation.