OpenMPI: Permission denied error while trying to use mpirun

问题

I would like to display "hello world" via MPI on different Google cloud compute instances with the help of the following code:

from mpi4py import MPI

size = MPI.COMM_WORLD.Get_size()
rank = MPI.COMM_WORLD.Get_rank()
name = MPI.Get_processor_name()

print("Hello, World! I am process/rank {} of {} on {}.\n".format(rank, size, name))

The problem is, that even so I can ssh-connect across all of these instances without problem, I get a permission denied error message when I try to run my script. I use following command to envoke my script:

mpirun --host localhost,instance_1,instance_2 python hello_world.py

And get the following error message:

Permission denied (publickey).
--------------------------------------------------------------------------
ORTE was unable to reliably start one or more daemons.
This usually is caused by:

* not finding the required libraries and/or binaries on
  one or more nodes. Please check your PATH and LD_LIBRARY_PATH
  settings, or configure OMPI with --enable-orterun-prefix-by-default

* lack of authority to execute on one or more specified nodes.
  Please verify your allocation and authorities.

* the inability to write startup files into /tmp (--tmpdir/orte_tmpdir_base).
  Please check with your sys admin to determine the correct location to use.

*  compilation of the orted with dynamic libraries when static are required
  (e.g., on Cray). Please check your configure cmd line and consider using
  one of the contrib/platform definitions for your system type.

* an inability to create a connection back to mpirun due to a
  lack of common network interfaces and/or no route found between
  them. Please check network connectivity (including firewalls
  and network routing requirements).
--------------------------------------------------------------------------

Additional information:

I installed open-MPI on all of my nodes
I have Google automatically set all of my ssh-keys by using gcloud to log into each instance from each instance
instance-type: n1-standard-1
instance-OS: Linux Debian (default)

Thanks you for your help :-)

New Information:
(thanks @ Zulan for pointing out that I should edit my previous post instead of creating a new answer for new information)

So, I tried to do the same with mpich instead of openmpi. However, I run into a similar error message.

Command:

mpirun --host localhost,instance_1,instance_2 python hello_world.py

Error message:

Host key verification failed.

I can ssh-connect between my two instances without problems, and through the gcloud commands the ssh-keys should automatically be set up properly.

So, has somebody an idea what the problem could be? I also checked the path, the firewall rules, and my ability to write startup scripts in the temp-folder. Can someone please try to recreate this problem? + Should I raise this question to Google? (never done such thing before, Im quite unsure :S)

Thanks for helping :)

回答1:

so I finally found a solution. Wow, problem was driving me nuts.

So it turned out, that I needed to generate ssh-keys manually for the script to work. I have no idea why, because google-services already set up the keys by using gcloud compute ssh , but well, it worked :)

Steps I did:

instance_1 $ ssh-keygen -t rsa
instance_1 $ cd .ssh
instance_1 $ cat id_rsa.pub >> authorized_keys
instance_1 $ gcloud compute copy-files id_rsa.pub 
instance_1 $ gcloud compute ssh instance_2

instance_2 $ cd .ssh
instance_2 $ cat id_rsa.pub >> authorized_keys

I will open another topic and ask why I cannot use ssh instance_2, even so gcloud compute ssh instance_2 is working. See: Difference between the commands "gcloud compute ssh" and "ssh"

来源：https://stackoverflow.com/questions/35819556/openmpi-permission-denied-error-while-trying-to-use-mpirun

标签

python

ssh

mpi

google-compute-engine

permission-denied