Open MPI logo

Open MPI User's Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Open MPI User's mailing list

Subject: Re: [OMPI users] Error running program : mca_oob_tcp_msg_send_handler: writev:failed: Bad file descriptor
From: Jeff Squyres (jsquyres) (jsquyres_at_[hidden])
Date: 2013-01-16 11:19:58


Make sure you disable iptables on all the nodes.

Also, check that all your IP interfaces are configured correctly. Do you have IP interfaces for only real ethernet connections and loopback? Or do you have other interfaces (e.g., for virtual machines)?

If you have interfaces for virtual machines, you'll need to exclude them from Open MPI -- see http://www.open-mpi.org/faq/?category=tcp#tcp-selection.

On Jan 16, 2013, at 10:59 AM, borja mf <borjamunozfernandez_at_[hidden]>
 wrote:

> Getting the same error...
> I forgot to say that I must to use Ubuntu and Im compiling with mpicc. My code is written on C.
>
> Thank for answer.
>
> Im going crazy with this problem. There's not much info about.
>
> 2013/1/16 Jeff Squyres (jsquyres) <jsquyres_at_[hidden]>
> Try disabling firewalling between your nodes. The easiest way is "sudo service iptables stop".
>
>
> On Jan 16, 2013, at 7:46 AM, borja mf <borjamunozfernandez_at_[hidden]>
> wrote:
>
> > Hello all.
> > I want to learn MPI and I've trying to setting up OMPI for first time on three nodes. My config above:
> > Ubuntu server - master node: pruebaborja
> > 2x Ubuntu Desktop - slaves node:
> > clienteprueba
> > clientepruebados 4 slots
> >
> > Im running NFSv4 for sharing /home/mpiuser.
> > I want to test a plain "Hello world"but I can't make it working successfully on node clienteprueba. This is the problem:
> >
> > mpiuser_at_pruebaborja:~$ mpirun -np 6 --hostfile .mpi_hostfile ./holamundo
> > [clienteprueba:01993] [[64434,0], 2] -> [[64434,0],0] mca_oob_tcp_msg_send_handler: writev:failed: Bad file descriptor (9) [sd = 9]
> > [clienteprueba:01993] [[64434,0], 2] routed:binomial: Connection to lifeline [[64434,0],0] lost
> >
> > However, with clientepruebados and pruebaborja only on my hostfile, it works:
> >
> > pruebaborja slots=1
> > clientepruebados slots=4
> > #clienteprueba slots=1
> >
> > mpiuser_at_pruebaborja:~$ mpirun -np 6 --hostfile .mpi_hostfile ./holamundo
> > Hola, mundo, soy pruebaborja: 0 de 6
> > Hola, mundo, soy pruebaborja: 5 de 6
> > Hola, mundo, soy clientepruebados: 1 de 6
> > Hola, mundo, soy clientepruebados: 2 de 6
> > Hola, mundo, soy clientepruebados: 3 de 6
> > Hola, mundo, soy clientepruebados: 4 de 6
> >
> > I've checked the OMPI versions on the machines and it's the same. I can't understand why Im getting this error on clienteprueba; i've done the same config on clientepruebados and clienteprueba. Anyone could help me to solve this?
> >
> > Sorry for my english.
> > Thanks in advance
> >
> >
> > _______________________________________________
> > users mailing list
> > users_at_[hidden]
> > http://www.open-mpi.org/mailman/listinfo.cgi/users
>
>
> --
> Jeff Squyres
> jsquyres_at_[hidden]
> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
>
>
> _______________________________________________
> users mailing list
> users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/users
>
> _______________________________________________
> users mailing list
> users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/users

-- 
Jeff Squyres
jsquyres_at_[hidden]
For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/