CODEWITHSUNDEEP
openmpi
user1607072
user1607072

Reputation: 141

"mpirun was unable to launch the specified application as it could not find an executable" error

I came across a strage problem, which worked but now it doesn't.

I run an OpenMPI program with tau profiling among 2 computers. It seems that mpirun can't run tau_exec program on a remote host, maybe it's a permission issue?

 cluster@master:~/software/mpi_in_30_source/test2$ mpirun -np 2 --hostfile hostfile -d           tau_exec -v -T MPI,TRACE,PROFILE ./hello.exe
 [master:19319] procdir: /tmp/openmpi-sessions-cluster@master_0/4568/0/0
 [master:19319] jobdir: /tmp/openmpi-sessions-cluster@master_0/4568/0
 [master:19319] top: openmpi-sessions-cluster@master_0
 [master:19319] tmp: /tmp
 [slave2:06777] procdir: /tmp/openmpi-sessions-cluster@slave2_0/4568/0/1
 [slave2:06777] jobdir: /tmp/openmpi-sessions-cluster@slave2_0/4568/0
 [slave2:06777] top: openmpi-sessions-cluster@slave2_0
 [slave2:06777] tmp: /tmp
 [master:19319] [[4568,0],0] node[0].name master daemon 0 arch ff000200
 [master:19319] [[4568,0],0] node[1].name slave2 daemon 1 arch ff000200
 [slave2:06777] [[4568,0],1] node[0].name master daemon 0 arch ff000200
 [slave2:06777] [[4568,0],1] node[1].name slave2 daemon 1 arch ff000200
 [master:19319] Info: Setting up debugger process table for applications
   MPIR_being_debugged = 0
   MPIR_debug_state = 1
   MPIR_partial_attach_ok = 1
   MPIR_i_am_starter = 0
   MPIR_proctable_size = 2
   MPIR_proctable:
     (i, host, exe, pid) = (0, master,      /home/cluster/software/mpi_in_30_source/test2/tau_exec, 19321)
     (i, host, exe, pid) = (1, slave2,      /home/cluster/software/mpi_in_30_source/test2/tau_exec, 0)
 --------------------------------------------------------------------------
 mpirun was unable to launch the specified application as it could not find an executable:

 Executable: tau_exec
 Node: slave2

 while attempting to start process rank 1.
 --------------------------------------------------------------------------
 [slave2:06777] sess_dir_finalize: job session dir not empty - leaving
 [slave2:06777] sess_dir_finalize: job session dir not empty - leaving
 [master:19319] sess_dir_finalize: job session dir not empty - leaving
 [master:19319] sess_dir_finalize: proc session dir not empty - leaving
 orterun: exiting with status -123

On slave2:

cluster@slave2:~/software/mpi_in_30_source/test2$ tau_exec -T MPI,TRACE,PROFILE ./hello.exe
hello MPI user: from process = 0 on machine=slave2, of NCPU=1 processes
cluster@slave2:~/software/mpi_in_30_source/test2$ which tau_exec
/home/cluster/tools/tau-2.22.2/arm_linux/bin/tau_exec

So there is a working tau_exec on both nodes. When I run mpirun without tau_exec everything works.

cluster@master:~/software/mpi_in_30_source/test2$ mpirun -np 2 --hostfile hostfile  ./hello.exe
hello MPI user: from process = 0 on machine=master, of NCPU=2 processes
hello MPI user: from process = 1 on machine=slave2, of NCPU=2 processes

Upvotes: 2

Views: 38683

Answers (4)

JobJob
JobJob

Reputation: 4127

If you're running a shell script with mpirun, make sure you've chmod +x script_file.sh else you'll see this error.

Upvotes: 2

wizardInCloud
wizardInCloud

Reputation: 159

Maybe is because you already had installed openMPI and not only MPICH2, so you should run the below commands as root:

root~# update-alternatives --config mpirun

There are 2 choices for the alternative mpirun (providing /usr/bin/mpirun).

Selection | Path | Priority | Status

  • *0 | /usr/bin/mpirun.openmpi | 50 | auto mode
  • 1 | /usr/bin/mpirun.mpich2 | 40 | manual mode
  • 2 | /usr/bin/mpirun.openmpi | 50 | manual mode

Press enter to keep the current choice[*], or type selection number: 1

Then you should select the MPICH version, as above, to run normally.

Upvotes: 2

Wissam Y. Khalil
Wissam Y. Khalil

Reputation: 153

once had an error like this when i tried to name the output file just try leave it the same

    mpirun -n <number> a.out

that is how it worked for me!

Upvotes: 2

Wesley Bland
Wesley Bland

Reputation: 9062

Try putting the full path to tau_exec in your command line. It's possible that you PATH isn't the same on all of the nodes. If that's the case, it wouldn't be able to find the executable anywhere where the path isn't correct.

It's most likely not a permission issue, but I don't remember all of the error messages in Open MPI to tell you how helpful they might be.

Upvotes: 2

Related Questions