[OMPI users] looking for serial implementation
I'm sure that I'm not the first person who wants their MPI program to compile when MPI is not available. It seems like the simplest solution to this is to have a header file (with implementation, or header file and .c file) that implements all of the functions for the case when MPI isn't available on a system. I found something called MPI_STUBS via google, but some of the links are broken, and I've searched the archives. This seems like the kind of thing that tons of people have done for themselves, and I'm hoping to avoid doing it for myself. Anybody have one they'd like to share? jgw -- John G. Wohlbier
Re: [OMPI users] looking for serial implementation
John Wohlbier wrote: I'm sure that I'm not the first person who wants their MPI program to compile when MPI is not available. It seems like the simplest solution to this is to have a header file (with implementation, or header file and .c file) that implements all of the functions for the case when MPI isn't available on a system. I found something called MPI_STUBS via google, but some of the links are broken, and I've searched the archives. This seems like the kind of thing that tons of people have done for themselves, and I'm hoping to avoid doing it for myself. Anybody have one they'd like to share? jgw I know this one (haven't tried it myself): http://wissrech.ins.uni-bonn.de/research/projects/nullmpi/ Regards, Dorian ___ users mailing list us...@open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users
[OMPI users] Problem with running openMPI program
I followed the steps given here to setup up openMPI cluster : http://www.ps3cluster.umassd.edu/step3mpi.html My cluster consists of two nodes, master(192.168.67.18) and salve(192.168.45.65), connected directly through a cross cable. After setting up the cluster n configuring the master node, i mounted /tmp folder of master node on the slave node(i had some problems with nfs at first but i worked my way out of it). Then i copied the 'pi.c' program in the /tmp folder and successfully complied it, giving me a binary file 'pi'. Now when i try to run the binary file using the following command #mpirun –np 2 ./Pi * * root@192.168.45.65's password: after entering the password it gives the following error: *bash: orted: command not found* *[ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file base/pls_base_orted_cmds.c at line 275* *[ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file pls_rsh_module.c at line 1166* *[ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file errmgr_hnp.c at line 90* *[ccomp.cluster:18963] ERROR: A daemon on node 192.168.45.65 failed to start as expected.* *[ccomp.cluster:18963] ERROR: There may be more information available from* *[ccomp.cluster:18963] ERROR: the remote shell (see above).* *[ccomp.cluster:18963] ERROR: The daemon exited unexpectedly with status 127.* *[ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file base/pls_base_orted_cmds.c at line 188* *[ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file pls_rsh_module.c at line 1198* *--* *mpirun was unable to cleanly terminate the daemons for this job. Returned value Timeout instead of ORTE_SUCCESS.* *--* * * I am totally lost now, as this is the first time i am working on a cluster project, and need some help Thank you Ankush
[OMPI users] Problem with insatlling OpenMPI on compute node
I followed the steps given here to setup up openMPI cluster : http://www.ps3cluster.umassd.edu/step3mpi.html My cluster consists of two nodes, master(running on fedora 10) and compute(CentOS 5.2) node, connected directly through a cross cable. I installed openmpi on both, there were many folders n files on de master node but there were none on de compute node. i again ran yum install openmpi successfully on de compute node but still there were no openmpi folders or files installed. Why is this happening? is it because of the OS?
Re: [OMPI users] Problem with running openMPI program
It might be best to: 1. Setup a non-root user to run MPI applications 2. Setup SSH keys between the hosts for this non-root user so that you can "ssh uptime" and not be prompted for a password/ passphrase This should help. On Apr 4, 2009, at 5:51 AM, Ankush Kaul wrote: I followed the steps given here to setup up openMPI cluster : http://www.ps3cluster.umassd.edu/step3mpi.html My cluster consists of two nodes, master(192.168.67.18) and salve(192.168.45.65), connected directly through a cross cable. After setting up the cluster n configuring the master node, i mounted /tmp folder of master node on the slave node(i had some problems with nfs at first but i worked my way out of it). Then i copied the 'pi.c' program in the /tmp folder and successfully complied it, giving me a binary file 'pi'. Now when i try to run the binary file using the following command #mpirun –np 2 ./Pi root@192.168.45.65's password: after entering the password it gives the following error: bash: orted: command not found [ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file base/ pls_base_orted_cmds.c at line 275 [ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file pls_rsh_module.c at line 1166 [ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file errmgr_hnp.c at line 90 [ccomp.cluster:18963] ERROR: A daemon on node 192.168.45.65 failed to start as expected. [ccomp.cluster:18963] ERROR: There may be more information available from [ccomp.cluster:18963] ERROR: the remote shell (see above). [ccomp.cluster:18963] ERROR: The daemon exited unexpectedly with status 127. [ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file base/ pls_base_orted_cmds.c at line 188 [ccomp.cluster:18963] [0,0,0] ORTE_ERROR_LOG: Timeout in file pls_rsh_module.c at line 1198 -- mpirun was unable to cleanly terminate the daemons for this job. Returned value Timeout instead of ORTE_SUCCESS. -- I am totally lost now, as this is the first time i am working on a cluster project, and need some help Thank you Ankush ___ users mailing list us...@open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users -- Jeff Squyres Cisco Systems
Re: [OMPI users] Problem with insatlling OpenMPI on compute node
I'm not too familiar with that tutorial nor your particular method of installation. In general, Open MPI needs a bunch of files to be available on all nodes (e.g., see if you can find "mpirun" on all nodes). See these FAQ entries: http://www.open-mpi.org/faq/?category=running#run-prereqs http://www.open-mpi.org/faq/?category=running#adding-ompi-to-path http://www.open-mpi.org/faq/?category=running#mpirun-prefix On Apr 4, 2009, at 7:23 AM, Ankush Kaul wrote: I followed the steps given here to setup up openMPI cluster : http://www.ps3cluster.umassd.edu/step3mpi.html My cluster consists of two nodes, master(running on fedora 10) and compute(CentOS 5.2) node, connected directly through a cross cable. I installed openmpi on both, there were many folders n files on de master node but there were none on de compute node. i again ran yum install openmpi successfully on de compute node but still there were no openmpi folders or files installed. Why is this happening? is it because of the OS? ___ users mailing list us...@open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users -- Jeff Squyres Cisco Systems
