Re: [HTCondor-users] single machine(slot) vanilla universe mpirun issue

[Jason Patton <jpatton@xxxxxxxxxxx>]

Hi Max,

It's strange that orte cares about the plm_rsh_agent when running on a single machine. A couple thoughts:

1. Try blanking the plm_rsh_agent parameter: --mca plm_rsh_agent ""

2. Sometimes the PATH inside a condor job might be missing or set improperly (maybe that's why it can't find ssh), could you try wrapping your executable in a script that does "export PATH" before running mpirun? You might even want to have your job run "env" and check your output file to see what kind of environment it's running in.

3. If your pool isn't set up to use a shared filesystem, condor mightÂbe transferring your mpirun executable. If your access point is at all different from your execution point, you might consider adding "transfer_executable = false" to your submit file to make sure you're using the mpirun executable on the execution point.

Let us know how it goes!

Jason Patton

On Fri, Sep 16, 2022 at 12:12 AM Zhen Song <snoopy1007@xxxxxxxxxxx> wrote:

Dear friends,

Issue:

I have 3 clusters already installed openmpi. The compiled mpi code works fine locally on each cluster. However, when I tried to use condor_submit, I got the following error:

--------------------------------------------------------------------------

                               âThe value of the MCA parameter "plm_rsh_agent" was set to a path

                               âthat could not be found:

                               â

                               â Âplm_rsh_agent: ssh : rsh

                               â

                               âPlease either unset the parameter, or check that the path is correct

âââââââââââââââââââââââââââââââââââââ

The submitting script is as follows, which I edited based onÂhttps://htcondor.readthedocs.io/en/latest/users-manual/parallel-applications.html#mpi-applications-within-htcondor-s-vanilla-universeÂ:

universe  = vanilla

executable = /usr/bin/mpirun

requestMemory = 1024

request_GPUs = 4

request_cpus = 4

arguments = -np 4 ./cmake_tmp/bin/main 0 1 500

log    Â= logs/job_$(Cluster).$(Process).log

output   = logs/job_$(Cluster).$(Process).out

error   Â= logs/job_$(Cluster).$(Process).error

should_transfer_files = yes

when_to_transfer_output = on_exit

transfer_input_files = ./cmake_tmp/bin/main

queue

Attempts:Â

I tried "arguments = -mca plm_rsh_agent /usr/lib/condor/libexec/condor_ssh -np 4 ./cmake_tmp/bin/main 0 1 500â. I got errors:

[huashan:596121] [[50260,0],0] ORTE_ERROR_LOG: Not found in file plm_rsh_module.c at line 231

[huashan:596121] [[50260,0],0] ORTE_ERROR_LOG: Not found in file ess_hnp_module.c at line 528

--------------------------------------------------------------------------

It looks like orte_init failed for some reason; your parallel process is

likely to abort. There are many reasons that a parallel process can

fail during orte_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):

 orte_plm_init failed

 --> Returned value Not found (-13) instead of ORTE_SUCCESS

--------------------------------------------------------------------------

Could anyone please help me sort it out?

Many thanks!

Best,

Max

_______________________________________________
HTCondor-users mailing list
To unsubscribe, send a message to htcondor-users-request@xxxxxxxxxxx with a
subject: Unsubscribe
You can also unsubscribe by visiting
https://lists.cs.wisc.edu/mailman/listinfo/htcondor-users

The archives can be found at:
https://lists.cs.wisc.edu/archive/htcondor-users/