Ponyo and Chihiro calculation clusters (by Loïc Marrec and Anne-Florence Bitbol, with the help of Benjamin Gallois, updated on 09/22/20) :
The LJP shares with the LKB and the LPTMC two computing clusters, called Ponyo and Chihiro. The following is a brief user guide. An official documentation is available at https://wiki-ponyo.lptmc.jussieu.fr/index.php/Accueil (available only via the Jussieu campus network or via a VPN). Be careful, the computing clusters do not replace your computer and are only to be used for long and/or heavy and/or parallelized codes. It is advisable to systematically test a code on your computer before using it on the clusters, in particular to know the required duration and memory (see point 6). Don't forget that clusters are shared by several researchers and that a misuse can penalize many people (who will surely point this out to you!). It is possible to use C, Python and Matlab on clusters, among others.
1- It is necessary to have an account to connect to clusters. If you don't have one, contact Michel Quaggetto (mq@lptl.jussieu.fr), research engineer at the LPTMC. Ask him explicitly to create an account for Ponyo and for Chihiro.
2- To connect to the Ponyo (respectively Chihiro) cluster from the Jussieu campus, open a terminal of your computer and enter the following command:
ssh USERNAME@ponyo.lptmc.jussieu.fr
(respectively ssh USERNAME@chihiro.lptmc.jussieu.fr
)
Enter your cluster access password and press the “Enter” key.
To connect to the Ponyo (respectively Chihiro) cluster from outside the Jussieu campus, open a terminal of your computer and enter the following command:
ssh -oProxyJump=ljp@134.157.132.110 USERNAME@ponyo.lptmc.jussieu.fr
(respectively ssh -oProxyJump=ljp@134.157.132.110 USERNAME@chihiro.lptmc.jussieu.fr
)
Enter the LJP server password (the same password as the LJP site password, with an additional “!”), press the “Enter” key, enter your cluster access password and press the “Enter” key.
Note: USERNAME is usually your last name. In case of doubt, refer to the document Cluster_Ponyo/Comptes_cluster.txt available on the LJP cloud, or contact Michel Quaggetto (mq@lptl.jussieu.fr), research engineer at the LPTMC.
3- At your first connection, change your password. To do so, connect to one of the clusters, Ponyo or Chihiro, and enter the following command:
passwd
(the initial password is normally aqwzsx99 for all)
Note: the password change will be effective for both clusters, Ponyo and Chihiro.
4- Once you are connected to one of the clusters, you can navigate in the folders using standard Unix commands, for example:
pwd
→ tells you which folder you are currently in
mkdir FOLDER
→ creates a folder called FOLDER in the current folder
cd FILE
→ takes you to the FILE folder
./FILE
→ executes the FILE file
chmod 755 FILE
→ change the permissions of the FILE file
etc.
5- Below, an example to copy files on the clusters, then to run a simple C code on one of the clusters (without any parallelization):
a- Download the Cluster_Ponyo/Simple_c folder available on the LJP cloud. Let's suppose that it is located at a PATH location on your computer.
b- To copy the folder “Simple_c” on the cluster from your computer, open a terminal of your computer and enter the following command:
scp -r PATH/Simple_c USERNAME@ponyo.lptmc.jussieu.fr:
Type your cluster access password and validate.
If you are not on the Jussieu campus, enter it instead:
scp -oProxyJump=ljp@134.157.132.110 -r CHEMIN/Simple_c USERNAME@ponyo.lptmc.jussieu.fr:
Type the password of the LJP server (the same password as the LJP site password, with a “!” in addition), validate, type your cluster access password and validate.
Note: all files copied to one cluster are automatically copied to the other. So, it doesn't matter if you type ponyo or chihiro during this step.
c- To check that the Simple_c folder is in your folder on the cluster, connect to one of the clusters, Ponyo or Chihiro, and run the command ls
. You should then see among the results “Simple_c”.
d- Go to the “Simple_c” folder and enter the following command:
cd Simple_c
e- Execute the bash script “RunCCode.sh” by entering the following command:
./RunCCode.sh
Note: in case of permission problem, give permission for execution by typing for example:
chmod 755 RunCCode.sh
f- The C code provided has normally been compiled and executed, and you should see an executable and result files appear in the “Simple_c” folder. Run ls
to see this.
g- You can now copy the result file from the cluster to your personal computer. To do this, open a terminal on your computer and type:
scp -r USERNAME@ponyo.lptmc.jussieu.fr:Simple_c/Times_s0.3_del0.1_K50_popNum1_mut3.00e-02_migr0.00e+00_1.dat .
If you are not on the Jussieu campus, enter instead:
scp -oProxyJump=ljp@134.157.132.110 -r USERNAME@ponyo.lptmc.jussieu.fr:Simple_c/Times_s0.3_del0.1_K50_popNum1_mut3.00e-02_migr0.00e+00_1.dat .
Don't forget the dot after the name of the file you want to copy. Check that the file “Times_s0.3_del0.1_K50_popNum1_mut3.00e-02_migr0.00e+00_1.dat” has now appeared in the home folder of your personal computer.
6- To adapt the above example to your code, you will need in particular to modify the CallCCode.sh file, which contains the instructions sent to the cluster (queue, time limit, memory requested, and for parallel programs, number of nodes and cores requested).
To do so, please refer to the following page of the official cluster documentation: https://wiki-ponyo.lptmc.jussieu.fr/index.php/Use_of_the_Slurm_batch_system.
On this page you will find the options available for the script you will use to submit a program to the cluster.
Thus, there are 4 possible “queues”, depending on the maximum execution time of your program :
def the default queue for jobs of less than 2 hours
p24 for jobs of less than 24 hours
p72 for jobs of less than 72 hours
p120 for jobs of less than 120 hours