The ABINIT Users Mailing List ( CLOSED )

["Ludwig, Christian" <ludwigc@uni-mainz.de>]

I just got another problem. On the master the parallel run works just fine, 
but when I want to start a job on a node from the master, I get

m_3016:  p4_error: rm_start: net_conn_to_listener failed: 14777
p0_11620:  p4_error: Child process exited while making connection to remote 
process on node: 0

When running from a node with 4 procs and with the option -np 4, I get 

Could not find enough machines for architecture LINUX

even though there should be.
When using -np 1, my log file looks like this
  ABINIT

  Give name for formatted input file:
abinit.in
  Give name for formatted output file:
abinit.out
  Give root name for generic input files:
abi
  Give root name for generic output files:
abo
  Give root name for generic temporary files:
tmp
-P-0000  leave_test : synchronization done...

and nothing more happens. The process idles at 0% CPU.
Especially this last thing confuses me, since it seems to work in principle, 
but for some reason it does not really do anything.

Any help is greatly appreciated.

Cheers,
Christian


________________________________________
Von: forum-owner@abinit.org [forum-owner@abinit.org] im Auftrag von Ludwig, 
Christian [ludwigc@uni-mainz.de]
Gesendet: Montag, 2. November 2009 15:47
An: forum@abinit.org
Betreff: AW: [abinit-forum] running MPI

Just a quick update. I found my own problem. mpich is using rsh, so having 
ssh set up properly did not help. I recompiled mpich with option -rsh=ssh, 
recompiled abinit and now it seems to work.


________________________________________
Von: forum-owner@abinit.org [forum-owner@abinit.org] im Auftrag von Ludwig, 
Christian [ludwigc@uni-mainz.de]
Gesendet: Mittwoch, 28. Oktober 2009 15:52
An: forum@abinit.org
Betreff: [abinit-forum] running MPI

Hello,

I have been working with a parallel abinit on an IBM p590 for a while. Since 
it is one machine with a bunch of processors, I just need to create a file 
machines with

localhost:4

to run a job on 4 processors. Then I start the job with

mpirun -np 4 -machinefile machines abinip < job.files >& log

This is all working and I wanted to try this on a x86 cluster. Before I 
continue, let me say that the master can ssh to itself and all nodes without 
password prompt.
I installed mpich and abinit on the master and the directories are mirrored 
to all nodes. When I execute mpirun on the master with a machinefile of

master:4

I get the error message

master: Connection refused
p0_3193:  p4_error: Child process exited while making connection to remote 
process on master: 0

written to the log file. With a machinefile of

node1:4

I get

Host address mismatch for 192.168.199.1
p0_3292:  p4_error: Child process exited while making connection to remote 
process on node1: 0

192.168.199.1 is not the ip of the host, I do not know how mpi gets the idea. 
Finally I tried to execute mpirun while being on node1 and got

Permission denied.
p0_28472:  p4_error: Child process exited while making connection to remote 
process on node1: 0


Hopefully one of you can give me a hint what to do to make it work.

Cheers,
Christian