how to run yade on multiple nodes using openmpi

Asked by xuanshenyu

Dear all,
 I want to run yade on multiple nodes using openmpi(4.0.3), however, I can find the correct command line in terminal to run yade.

I have tried the command "mpirun -np 120 -host node3:128 yade case.py", and there would send an error message:"bash: orted: command not foud"

Ubuntu:20.04
Yade:2022.01a

Question information

Language:
English Edit question
Status:
Answered
For:
Yade Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
Jérôme Duriez (jduriez) said :
#1

Hi,

While I'm not sure of the answer to your title question, solving "bash: orted: command not foud" should easily be solved with a proper system installation for openmpi simulations (that would include openmpi-bin package in particular)

Revision history for this message
xuanshenyu (shenyuxuan) said :
#2

Hi,
Thanks for kindly reply, I have solved the question of openmpi ( "bash: orted: command not foud").
And I carry out a test command of openmpi, "root@xsy: mpirun --allow-run-as-root --host master:128 --prefix /disk1/openmpi4.1.3 -n 4 hostname" and there were feedback :
master
master
master
master

However, when I carry out "root@xsy: mpirun --allow-run-as-root --host master:128 --prefix /disk1/openmpi4.1.3 -n 4 yade-2022.1a helloMPI.py", there were feedback:

3 Open MPI tried to fork a new process via the "execve" system call but
4 failed. Open MPI checks many things before attempting to Launch a
5 child process, but nothing is perfect. This error may be indicative
6 of another problem on the target host, or even something as silly as
7 having specified a directory for your application. Your job will now
8 abort.
9
10 Local host: master
11 Working dir: /disk1/yade/16/install/bin
12 Application name: /disk1/yade/16/install/bin/yade-2022.01a
13 Еггог: /disk1/yade/16/install/bin/yade-2022.01a has a bad interpreter on the first line.
14 ------------------------------------
15------------------------------------
16 mpirun was unable to start the specified application as it encountered an
17 error:
18
19 Error code: 1
20 Error name: (null)
21 Node: master
22
23 when attempting to start process rank 0.
24 -----------------------------
25 4 total processes failed to start
26 [xsy:023661 3 more processes have sent help message help-orte-odls-default.txt / execve error
27 [xsy:02366] Set MCA parameter "orte_ base_ help_aggregate" to o to see all help / error messages
28 root@sy: /disk1/yade/16/install/bin#

Revision history for this message
Bruno Chareyre (bruno-chareyre) said :
#3

Hi,
Your problems seem very platform dependent.
I have been working in a couple different contexts and I have never seen a need for the flag " --allow-run-as-root".

I would suggest you try running simple parallel jobs.
"hostname" is a good start but it is excessively simple. Try running a python script for instance, then if it works try an example script from the mpi4py module (i.e. an actual parallel job), then only if the previous two work it could be that yade's implementation has an issue on you specific system (in this case please report here).

Bruno

Revision history for this message
xuanshenyu (xuanshenyu) said :
#4

Hi, Bruno
I was very encouraged by your professional reponse.

>> 1. I have never seen a need for the flag " --allow-run-as-root".
       Yes, my platform context is very special, because the systerm of cluster is Centos.7 and off-line in our school, and we failed to install Yade on Centos.7. So, my cluster engineer suggest that we use an ubuntu laptop to call the cluster.
       After a series of setting such as group lan, encryption-free login, encryption-free communication, we have to use root perimission to perform parallel computing.
>>2. Try running a python script for instance, then if it works try an example script from the mpi4py module (i.e. an actual parallel job)...
        That's a good idea. Could you a more detail suggestion or a python script for MPI? I know a little of MPI and mpi4py.
        I'm sorry if this caused you any stress.
Thanks for your help.

Can you help with this problem?

Provide an answer of your own, or ask xuanshenyu for more information if necessary.

To post a message you must log in.