Problemas com o mpirun (openmpi)

Question

Eu instalei localmente duas versões openmpi no meu cluster:

• openmpi-1.8.1: quando eu executo o mpirun sob esta versão, isso me dá um erro:

  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  librdmacm: Fatal: unable to open RDMA device
  

• openmpi-2.0.0: quando eu executo o mpirun sob esta versão, ele me diz:

  mca_base_component_repository_open: shmem "/opt/openmpi-1.8.1/lib/openmpi/mca_shmem_posix" uses an MCA interface that is not recognized (component MCA v2.0.0 != supported MCA v2.1.0) -- ignored
  
  It looks like opal_init failed for some reason; your parallel process is
  likely to abort.  There are many reasons that a parallel process can
  fail during opal_init; some of which are due to configuration or
  environment problems.  This failure appears to be an internal failure;
  here's some additional information (which may only be relevant to an
  Open MPI developer):
    opal_shmem_base_select failed
    --> Returned value -1 instead of OPAL_SUCCESS
  

Ambas as versões são instaladas localmente em / opt e carregadas como módulos. Parece que o openmpi-2.0.0 ainda usa dependências do openmpi-1.8.1, que eu não entendo.

Eu apreciarei quaisquer diagnósticos e / ou dicas para resolvê-lo.

Obrigado antecipadamente.

    

Answer 1

Você não deve instalá-los no mesmo local em que suas bibliotecas estão em conflito. O openmpi 2.0.0 usa a interface MCA 2.1.0, o openmpi 1.8.1 usa a interface MCA 2.0.0. Então você deve instalar as bibliotecas em locais separados