http://blog.gmane.org/gmane.comp.clustering.open-mpi.user hourly 1 1901-01-01T00:00+00:00 http://gmane.org/img/gmane-25t.png http://gmane.org http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19141 <pre>Dear users, I have been banging my head against the wall for some time to find a reliable and portable way to determine if a call to MPI::File::Open() was successful or not. Let me give some background information first. We develop an open-source astrophysical modeling code called Cloudy. This is used by many scientists on a variety of platforms. We obviously have no control over the MPI version that is installed on that platform, it may not even be open-MPI. So what we need is a method that is supported by all MPI distros. Our code is written in C++, so we use the C++ version of the MPI and MPI-IO libraries. Any help would be greatly appreciated. Cheers, Peter. </pre> Peter van Hoof 2013-05-17T09:00:36 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19130 <pre>Dear Support, We are having an issue with our OMPI runs. When we run jobs on &lt;=550 machines (550 x 16 cores) then they work without any problem. As soon as we run them on 600 or more machines we get the "plm:tm: failed to spawn daemon, error code = 17000" Error We are using: OpenMPI ver: 1.6.4 (Compiled with GCC v4.4.6) Torque ver: 2.5.12 The ompi_info's output is attached. The Environmentstats have been pasted below. Please assist. env envsubst [ocfacc&lt; at &gt;cyan01 fullrun]$ env MODULE_VERSION_STACK=3.2.10 OMPI_MCA_mtl=^psm MANPATH=/local/software/openmpi/1.6.4/gcc/share/man:/local/software/moab/6.1.10/man:/usr/local/share/man:/usr/share/man/overrides:/usr/share/man:/local/Modules/default/share/man HOSTNAME=cyan01 SHELL=/bin/bash TERM=xterm HISTSIZE=1000 QTDIR=/usr/lib64/qt-3.3 OLDPWD=/home/ocfacc/hpl/fullrun/results QTINC=/usr/lib64/qt-3.3/include LC_ALL=POSIX USER=ocfacc LD_LIBRARY_PATH=/local/software/openmpi/1.6.4/gcc/lib:/local/software/torque/default/lib LS_COLORS=rs=0:di=01;34:ln=01;36:m</pre> Qamar Nazir 2013-05-16T16:09:19 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19129 <pre> I am having trouble configuring OpenMPI-1.6.4 with the Intel C/C++ composer (12.0.2). My OS is OSX 10.7.5. I am not a computer whizz so I hope I can explain what I did properly: 1) In bash, I did source /opt/intel/bin/compilervars.sh intel64 and then echo PATH showed: /opt/intel/composerxe-2011.2.142/bin/intel64:/opt/intel/composerxe-2011.2.142/mpirt/bin/intel64:/opt/intel/composerxe-2011.2.142/bin:/Library/Frameworks/EPD64.framework/Versions/Current/bin:/Library/Frameworks/Python.framework/Versions/Current/bin:.:/Library/Frameworks/EPD64.framework/Versions/Current/bin:/usr/bin:/bin:/usr/sbin:/sbin:/usr/local/bin:/usr/X11/bin 2) which icc and which icpc showed: /opt/intel/composerxe-2011.2.142/bin/intel64/icc and /opt/intel/composerxe-2011.2.142/bin/intel64/icpc So that all seems okay to me. Still when I do ./configure CC=icc CXX=icpc F77=ifort FC=ifort --prefix=/opt/openmpi-1.6.4 from the folder in which the extracted OpenMPI files sit, I get =========================================================</pre> Geraldine Hochman-Klarenberg 2013-05-16T15:57:42 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19126 <pre>Hi Do we need distributed file system (like NFS) when running MPI program on multiple machines? thanks, Reza _______________________________________________ users mailing list users&lt; at &gt;open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users</pre> Reza Bakhshayeshi 2013-05-16T15:24:41 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19120 <pre> Dear mpi team / users: To get a mpi with 64-bit integers (linux system: ubuntu 12.04) I invoked the following configuration options: ./configure --prefix=/opt/openmpi CXX=icpc CC=icc F77=ifort FC=ifort FFLAGS=-i8 FCFLAGS=-i8 The subsequent make/install scripts apparently went through smoothly, but when I check ompi_info -a | grep 'Fort integer size' the result reads: Fort integer size: 4 What went awry? For all hints and suggestions many thanks in advance, Hans H. </pre> H Hogreve 2013-05-15T23:42:48 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19112 <pre>I'm been playing with come code to try and become familiar with MPI_Comm_accept and MPI_Comm_connect to implement an MPI client/server. The code that I have simply sends a single MPI_INT, the client process pid, to the server and then disconnects. The code that I have works for a few test runs but then on the 2nd or 3rd client connection, the server seems to stop responding and the client spins 100% CPU in the call to MPI_Comm_accept. Am I doing something wrong in my code? Thanks in advance for any help. First, an example run ... In terminal #1, start the name service $ ompi-server -r ${PREFIX}/var/run/ompi-server/uri.txt $ In terminal #2, start the server code $ mpirun -mca btl tcp,sm,self \ --ompi-server file:${PREFIX}/var/run/ompi-server/uri.txt mpi-server mpi-server pid 41556 Opened port 2011758592.0;tcp://10.161.1.73:51113+2011758593.0;\ tcp://10.161.1.73:51114:300 MPI_Info_set("ompi_global_scope", "true") Published {"mpi-server-example", "2011758592.0;\ tcp://10.161.1.73:51113+2011758593.0;tcp</pre> Damien Kick 2013-05-14T18:15:31 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19110 <pre>Hello I'm kuni255 I build bewulf type PC Cluster (Cent OS release 6.4). And I studing about MPI.(Open MPI Ver.1.6.4) I tried following sample which using MPI_REDUCE. Then, Error occured. This cluster system consist of one head node and 2 slave nodes. And sharing home directory in head node by NFS. so Open MPI is installed each nodes. When I test this program on only head node, program is run correctly. and output result. But When I test this program on only slave node, same error occured. Please tell me, good idea : ) Error message [bwslv01:30793] *** An error occurred in MPI_Reduce: the reduction operation MPI_SUM is not defined on the MPI_INTEGER datatype [bwslv01:30793] *** on communicator MPI_COMM_WORLD [bwslv01:30793] *** MPI_ERR_OP: invalid reduce operation [bwslv01:30793] *** MPI_ERRORS_ARE_FATAL: your MPI job will now abort -------------------------------------------------------------------------- mpirun has exited due to process rank 1 with PID 30793 on node bwslv01 exiting improperly. There a</pre> Hayato KUNIIE 2013-05-14T15:39:06 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19106 <pre>Hello All,I am new user to this mailing list and am trying to get familiar with mpi.I am using open mpi version 1.4.3I have a simple shell script that I have writtenThe Shell Script that I have is as followshelloWorld.sh&lt;code&gt;#!/bin/bashecho " Hello World from Rank $rank"&lt;code&gt; The MPI Command that I am executing is as followsmpirun -np 3 helloWorld.sh What I want to know how what parameter do I need to pass from the mpirun command line so that I can read the rank for each task within my helloWorld Shell Script. Please excuse me if this is a duplicate entry. I have searched through all the forums and did not get the answer to my question.Thanks in advance for all your help.Regards,Deepak _______________________________________________ users mailing list users&lt; at &gt;open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users</pre> deepak mehta 2013-05-10T20:39:34 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19098 <pre>Dear All, I have recently installed gcc 4.7.3 on my cent OS 6.4. Moreover, I have compiled openmpi 1.6.4 with the above compiler. My LD_LIBRARY_PATH is set correctly and it points to both /lib and /lib64 where libgfortran.so and libgcc_s.so for 32 and 64 bits are located. Every time I compile a Fortran, a C or a C++ source code with the wrapper mpif90, mpicc or mpicx I get this warning: /usr/bin/ld: skipping incompatible /scratch/home0/pmatteo /research/lib_install/lib/libgfortran.so when searching for -lgfortran or /usr/bin/ld: skipping incompatible /scratch/home0/pmatteo /research/lib_install/lib/libgcc_s.so when searching for -lgcc_s I have switch my LD_LIBRARY_PATH as suggested in this thread: http://www.open-mpi.org/community/lists/users/2009/02/8067.php but nothing change. Any idea what I am doing wrong? I know that it is just a warning but I would like to avoid it. Thank you. </pre> Matteo Parsani 2013-05-08T19:12:55 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19093 <pre>Are people in community using MPI libraries in their application for real time processing / analytics? I have access to a cluster with infiniband and want to test millions of hypotheses and depending upon which one passed would run an aggregate function over them. I would like to achieve a subsecond response for this and was wondering how one could take care of the network calls overhead? Thanks! _______________________________________________ users mailing list users&lt; at &gt;open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users</pre> RoboBeans 2013-05-07T17:48:44 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19083 <pre>Hello everyone, I'm having troubles with building OpenMPI v1.6.4 (actually all v1.6 and v1.7 versions) release sources (in both x86 and x64 modes) on Windows machine using VS 2010 and VS 2012 compilers and cmake v2.8.10.2. Just to note, I tried it with VS2008 too and it worked fine in x86 mode, but failed in x64 mode. Here is error I get with VS2010 and VS2012 when I execute 'cmake-gui ..' and push Configure: CMake Error at C:/Program Files (x86)/CMake 2.8/share/cmake-2.8/Modules/CMakeFortranInformation.cmake:27 (get_filename_component): get_filename_component called with incorrect number of arguments Call Stack (most recent call first): contrib/platform/win32/CMakeModules/setup_f77.cmake:18 (include) contrib/platform/win32/CMakeModules/ompi_configure.cmake:616 (INCLUDE) CMakeLists.txt:99 (BEGIN_CONFIGURE) If I naively mask off using setup_F77 from ompi_configure.cmake (because I don't wan't Fortran support), I get following compile error: 1&gt;------ Build started: Project: libmpi_cxx, Configuration: Debug</pre> Nenad Vujicic 2013-05-07T15:16:37 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19076 <pre>Hi everyone, I want to install OpenMPI on LSF cluster in our organization. I am not proficient with Linux/LSF, and some of my questions might be from lack of understanding of the system, and not related to OpenMPI directly. So far I found these bits of information on the site of OpenMPI *1.* *OpenMPI support for 1.6 seems to be broken, and was fixed maybe in 1.7?* http://www.open-mpi.org/community/lists/users/2013/03/21640.php *2. The installation on LSF is supposed to be easy:* http://www.open-mpi.org/faq/?category=building#build-rte-lsf My questions are: * * *1*. What is the latest stable version that is known to integrate in a native way with LSF? *2*. When building with LSF support, in what directory should I run the ./configure and makes scripts from? Should I be logined into one of the hosts of LSF cluster? *3*. Will these scripts copy openmpi shared libraries into each host on the cluster? *4*. Where will the mpi compiler be after the installation? What include pathes and libraries should I add </pre> Andrey Rubshtein 2013-05-07T14:09:42 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19071 <pre>Hi folks, I am testing our cluster with module environment, and am having a headache to understand openmpi 1.7.2!!! So our system currently has openmpi 1.6.3 (at default location /usr/local), 1.6.4 and 1.7.2 compiled with intel compilers (installed at /opt/apps). In order to use openmpi 1.7.2 for example, I tried: $ module load mpi/openmpi-1.7.2_composer_xe_2013.3.163 $ module load apps/abinit-7.2.1_composer_xe_2013.3.163_openmpi-1.7.2_intel_fftw3-mkl $ mpirun ./mpihello_intel mca: base: component_find: unable to open /usr/local/lib/openmpi/mca_ess_hnp: /usr/local/lib/openmpi/mca_ess_hnp.so: undefined symbol: orte_local_jobdata (ignored) mca: base: component_find: unable to open /usr/local/lib/openmpi/mca_ess_slurm: /usr/local/lib/openmpi/mca_ess_slurm.so: undefined symbol: orte_orted_exit_with_barrier (ignored) mca: base: component_find: unable to open /usr/local/lib/openmpi/mca_ess_slurmd: /usr/local/lib/openmpi/mca_ess_slurmd.so: undefined symbol: orte_pmap_t_class (ignored) mpirun: symbol lookup error:</pre> Duke Nguyen 2013-05-07T11:36:39 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19064 <pre>Hi, I have used OpenMPI before without any troubles, and configured MPICH, MPICH2 and OpenMPI in many different machines before, but recently we upgraded the OS to Fedora 17, and now I'm having trouble running an MPI code in two of our machines connected via a switch. I thought perhaps the old installation was giving problems, so I reinstalled OpenMPI (1.6.4) and I have no trouble when running a parallel code in just one node. I also don't have any trouble ssh'ing (without need for password) between these machines, but when I try to run a parallel job spanning both machines, I get a hanged mpiexec process in the submitting machine, and an "orted" process in the other machine, but nothing moves. I guess it is an issue with libraries and/or different MPI versions (the machines have other site-wide MPI libraries installed), but I'm not sure how to debug the issue. I looked in the FAQ, but I didn't find anything relevant. Issue http://www.open-mpi.org/faq/?category=running#intel-compilers-static is different</pre> Angel de Vicente 2013-05-04T23:54:00 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19063 <pre>Hello, when installing openmpi-1.6.4 (`cd /root/openmpi-1.6.4; ./configure --prefix=/usr/opt/openmpi-1.6.4/gcc-4.7.2/mxib --with-openib --with-mx=/opt/mx/1.2.16/3.2.0-4-amd64 --with-slurm; make; make install`) we found some references to non-existing paths in *.la files as in the following example: `cat /usr/opt/openmpi-1.6.4/gcc-4.7.2/mxib/lib/libmpi.la` (this is the installed version of openmpi after the `make install` from above): ... dependency_libs=' -L/root/openmpi-1.6.4/opal/mca/hwloc/hwloc132/hwloc/src -ldl -lrt -lnsl -lutil -lm' ... Now applications using GNU libtool will stop here during their compilation with an error because this path (which is a subdirectory of the path were I compiled openmpi on a different system!) does not exist. (And even if it would exist this would lead to a "permission denied" error because /root is not readable ...and finally there are no libs in the src directory /root/openmpi-1.6.4/opal/mca/hwloc/hwloc132/hwloc/src). # are the *.la files necessary at all? (probabl</pre> Stefan Friedel 2013-05-03T14:19:03 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19060 <pre>If I'm using MPI_Send(...) and MPI_Recv(...) in a producer/consumer model and choose not to buffer messages internally (in the app), allowing them to acumulate in the MPI layer, how large of an MPI message queue can I expect before something breaks? ---John </pre> John Chludzinski 2013-05-02T19:23:10 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19054 <pre>Hi, I have a program where each MPI rank hosts a set of data blocks. After doing computation over *some of* its local data blocks, each MPI rank needs to exchange data with other ranks. Note that the computation may involve only a subset of the data blocks on a MPI rank. The data exchange is achieved at each MPI rank through Isend and Irecv and then Waitall to complete the requests. Each pair of Isend and Irecv exchanges a corresponding pair of data blocks at different ranks. Right now, we do Isend/Irecv for EVERY block! The idea is that because the computation at a rank may only involves a subset of blocks, we could mark those blocks as dirty during the computation. And to reduce data exchange bandwidth, we could only exchanges those *dirty* pairs across ranks. The problem is: if a rank does not compute on a block 'm', and if it does not call Isend for 'm', then the receiving rank must somehow know this and either a) does not call Irecv for 'm' as well, or b) let Irecv for 'm' fail gracefully. My questi</pre> Thomas Watson 2013-05-01T17:28:26 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19046 <pre>Hello, My recent job started normally but after a few hours of running died with the following message: -------------------------------------------------------------------------- A daemon (pid 19390) died unexpectedly with status 137 while attempting to launch so we are aborting. There may be more information reported by the environment (see above). This may be because the daemon was unable to find all the needed shared libraries on the remote node. You may set your LD_LIBRARY_PATH to have the location of the shared libraries on the remote nodes and this will automatically be forwarded to the remote nodes. -------------------------------------------------------------------------- -------------------------------------------------------------------------- mpirun noticed that the job aborted, but has no info as to the process that caused that situation. The scheduling script is below: #$ -S /bin/bash #$ -cwd #$ -N SC3blastx_64-96thr #$ -pe openmpi* 64-96 #$ -l h_rt=24:00:00,vf=3G </pre> Vladimir Yamshchikov 2013-04-30T19:26:53 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19038 <pre>I have a number of processes split into sender and receivers. Senders read large quantities of randomly organised data into buffers for transmission to receivers. When a buffer is full it needs to be transmitted to all receivers this repeats until all the data is transmitted. Problem is that MPI_Bcast must know the root it is to receive from and therefore cant receive 'blind' from the first full sender. Scatter would be inneffieienct because a few senders wont have anything to send - so its wasteful to transmit those empty buffers repeatedly.  Any ideas? Can Bcast recievers be promiscuous? Thanks Randolph_______________________________________________ users mailing list users&lt; at &gt;open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users</pre> Randolph Pullen 2013-04-30T06:43:23 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19033 <pre>Dear users, I am getting following error while doing a calculation. The job is getting terminated before writing anything in output file . ========================================================================================== ssh: ibc18: Name or service not known^M -------------------------------------------------------------------------- A daemon (pid 8103) died unexpectedly with status 255 while attempting to launch so we are aborting. There may be more information reported by the environment (see above). This may be because the daemon was unable to find all the needed shared libraries on the remote node. You may set your LD_LIBRARY_PATH to have the location of the shared libraries on the remote nodes and this will automatically be forwarded to the remote nodes. -------------------------------------------------------------------------- -------------------------------------------------------------------------- mpirun noticed that the job aborted, but has no info as to the process that caused that situ</pre> sudhirs< at > 2013-04-29T13:12:50 http://comments.gmane.org/gmane.comp.clustering.open-mpi.user/19029 <pre>Hi, I'm new on this list. I'm using MPI for years but I don't have written a lot of code with MPI. Therefore is my question perhaps ridiculous: I'm using a Computational Fluid Mechanics (CFD) Solver. This Solver uses MPI to exchange the data between the different partitions. In this solver the "root processor" is always the processor 1. So this proc reads the input, broadcast a lot of things and writes the output. During a time step the solver computes the reference pressure at a point. This computation is done on a processor, which may not be the root processor. Therefore after the computation a broadcast of the value is necessary. For the moment in the code the broadcast is done with the processor, where the reference pressure is computed, as root processor (and not with the standard "root processor"). Is it false ? Must the root processor be the same during a computation for all broadcasts ? Best regards, Guillaume </pre> giggzounet 2013-04-29T11:15:04 Search the mailing list at Gmane query http://search.gmane.org/?group=$group=gmane.comp.clustering.open-mpi.user