Re: [OMPI users] Fwd: Fwd: problem in cluster
That fine. But in your job script ppn=2 Also check ldd cgles on the compute servers themselves. Are all the libraries available in your path? On 25 April 2018 at 11:43, Ankita m wrote: > i have 16 cores per one node. I usually use 4 node each node has 16 cores > so total 64 processes. > > On Wed, Apr 25, 2018 at 2:57 PM, John Hearns via users < > users@lists.open-mpi.org> wrote: > >> I do not see much wrong with that. >> However nodes=4 ppn=2 makes 8 processes in all. >> You are using mpirun -np 64 >> >> Actually it is better practice to use the PBS supplied environment >> variables during the job, rather than hard-wiring 64 >> I dont have access to a PBS cluster from my desk at the moment. >> You could also investigate using mpiprocs=2 Then I think with openmpi >> if it has compiled in PBS support all you would have to do is >> mpirun >> >> Are you sure your compute servers only have two cores ?? >> >> I also see that you are commenting out the module load openmpi-3.0.1 I >> would guess you want the default Opnempi, which is OK >> >> First thing I would do, before the mpirun line in that job script: >> >> which mpirun(check that you are picking up an Openmpi version) >> >> ldd ./cgles (check you are bringing in the libraries that you should) >> >> >> Also run mpirun with the verbose flag -v >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> On 25 April 2018 at 11:10, Ankita m wrote: >> >>> while using openmpi- 1.4.5 the program ended by showing this error file (in the attachment) >>> >>> I am Using PBS file . Below u can find the script that i am using to >>> run my program >>> >>> ___ >>> users mailing list >>> users@lists.open-mpi.org >>> https://lists.open-mpi.org/mailman/listinfo/users >>> >> >> >> ___ >> users mailing list >> users@lists.open-mpi.org >> https://lists.open-mpi.org/mailman/listinfo/users >> > > > ___ > users mailing list > users@lists.open-mpi.org > https://lists.open-mpi.org/mailman/listinfo/users > ___ users mailing list users@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/users
Re: [OMPI users] Fwd: Fwd: problem in cluster
Ankita, please read here:https://www.open-mpi.org/faq/?category=mpi-apps On 25 April 2018 at 11:44, Ankita m wrote: > Can you please tell me whether to use mpicc compiler ar any other compiler > for openmpi programs > > On Wed, Apr 25, 2018 at 3:13 PM, Ankita m wrote: > >> i have 16 cores per one node. I usually use 4 node each node has 16 cores >> so total 64 processes. >> >> On Wed, Apr 25, 2018 at 2:57 PM, John Hearns via users < >> users@lists.open-mpi.org> wrote: >> >>> I do not see much wrong with that. >>> However nodes=4 ppn=2 makes 8 processes in all. >>> You are using mpirun -np 64 >>> >>> Actually it is better practice to use the PBS supplied environment >>> variables during the job, rather than hard-wiring 64 >>> I dont have access to a PBS cluster from my desk at the moment. >>> You could also investigate using mpiprocs=2 Then I think with openmpi >>> if it has compiled in PBS support all you would have to do is >>> mpirun >>> >>> Are you sure your compute servers only have two cores ?? >>> >>> I also see that you are commenting out the module load openmpi-3.0.1 I >>> would guess you want the default Opnempi, which is OK >>> >>> First thing I would do, before the mpirun line in that job script: >>> >>> which mpirun(check that you are picking up an Openmpi version) >>> >>> ldd ./cgles (check you are bringing in the libraries that you should) >>> >>> >>> Also run mpirun with the verbose flag -v >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> >>> On 25 April 2018 at 11:10, Ankita m wrote: >>> > while using openmpi- 1.4.5 the program ended by showing this error > file (in the attachment) > I am Using PBS file . Below u can find the script that i am using to run my program ___ users mailing list users@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/users >>> >>> >>> ___ >>> users mailing list >>> users@lists.open-mpi.org >>> https://lists.open-mpi.org/mailman/listinfo/users >>> >> >> > > ___ > users mailing list > users@lists.open-mpi.org > https://lists.open-mpi.org/mailman/listinfo/users > ___ users mailing list users@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/users
Re: [OMPI users] Fwd: Fwd: problem in cluster
Can you please tell me whether to use mpicc compiler ar any other compiler for openmpi programs On Wed, Apr 25, 2018 at 3:13 PM, Ankita m wrote: > i have 16 cores per one node. I usually use 4 node each node has 16 cores > so total 64 processes. > > On Wed, Apr 25, 2018 at 2:57 PM, John Hearns via users < > users@lists.open-mpi.org> wrote: > >> I do not see much wrong with that. >> However nodes=4 ppn=2 makes 8 processes in all. >> You are using mpirun -np 64 >> >> Actually it is better practice to use the PBS supplied environment >> variables during the job, rather than hard-wiring 64 >> I dont have access to a PBS cluster from my desk at the moment. >> You could also investigate using mpiprocs=2 Then I think with openmpi >> if it has compiled in PBS support all you would have to do is >> mpirun >> >> Are you sure your compute servers only have two cores ?? >> >> I also see that you are commenting out the module load openmpi-3.0.1 I >> would guess you want the default Opnempi, which is OK >> >> First thing I would do, before the mpirun line in that job script: >> >> which mpirun(check that you are picking up an Openmpi version) >> >> ldd ./cgles (check you are bringing in the libraries that you should) >> >> >> Also run mpirun with the verbose flag -v >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> >> On 25 April 2018 at 11:10, Ankita m wrote: >> >>> while using openmpi- 1.4.5 the program ended by showing this error file (in the attachment) >>> >>> I am Using PBS file . Below u can find the script that i am using to >>> run my program >>> >>> ___ >>> users mailing list >>> users@lists.open-mpi.org >>> https://lists.open-mpi.org/mailman/listinfo/users >>> >> >> >> ___ >> users mailing list >> users@lists.open-mpi.org >> https://lists.open-mpi.org/mailman/listinfo/users >> > > ___ users mailing list users@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/users
Re: [OMPI users] Fwd: Fwd: problem in cluster
i have 16 cores per one node. I usually use 4 node each node has 16 cores so total 64 processes. On Wed, Apr 25, 2018 at 2:57 PM, John Hearns via users < users@lists.open-mpi.org> wrote: > I do not see much wrong with that. > However nodes=4 ppn=2 makes 8 processes in all. > You are using mpirun -np 64 > > Actually it is better practice to use the PBS supplied environment > variables during the job, rather than hard-wiring 64 > I dont have access to a PBS cluster from my desk at the moment. > You could also investigate using mpiprocs=2 Then I think with openmpi if > it has compiled in PBS support all you would have to do is > mpirun > > Are you sure your compute servers only have two cores ?? > > I also see that you are commenting out the module load openmpi-3.0.1 I > would guess you want the default Opnempi, which is OK > > First thing I would do, before the mpirun line in that job script: > > which mpirun(check that you are picking up an Openmpi version) > > ldd ./cgles (check you are bringing in the libraries that you should) > > > Also run mpirun with the verbose flag -v > > > > > > > > > > > > > > > > > > > > > > > > > > > > > On 25 April 2018 at 11:10, Ankita m wrote: > >> >>> while using openmpi- 1.4.5 the program ended by showing this error file >>> (in the attachment) >>> >> >> I am Using PBS file . Below u can find the script that i am using to run >> my program >> >> ___ >> users mailing list >> users@lists.open-mpi.org >> https://lists.open-mpi.org/mailman/listinfo/users >> > > > ___ > users mailing list > users@lists.open-mpi.org > https://lists.open-mpi.org/mailman/listinfo/users > ___ users mailing list users@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/users
Re: [OMPI users] Fwd: Fwd: problem in cluster
I do not see much wrong with that. However nodes=4 ppn=2 makes 8 processes in all. You are using mpirun -np 64 Actually it is better practice to use the PBS supplied environment variables during the job, rather than hard-wiring 64 I dont have access to a PBS cluster from my desk at the moment. You could also investigate using mpiprocs=2 Then I think with openmpi if it has compiled in PBS support all you would have to do is mpirun Are you sure your compute servers only have two cores ?? I also see that you are commenting out the module load openmpi-3.0.1 I would guess you want the default Opnempi, which is OK First thing I would do, before the mpirun line in that job script: which mpirun(check that you are picking up an Openmpi version) ldd ./cgles (check you are bringing in the libraries that you should) Also run mpirun with the verbose flag -v On 25 April 2018 at 11:10, Ankita m wrote: > >> while using openmpi- 1.4.5 the program ended by showing this error file >> (in the attachment) >> > > I am Using PBS file . Below u can find the script that i am using to run > my program > > ___ > users mailing list > users@lists.open-mpi.org > https://lists.open-mpi.org/mailman/listinfo/users > ___ users mailing list users@lists.open-mpi.org https://lists.open-mpi.org/mailman/listinfo/users