[OMPI users] Open-MPI and TCP port range
Hi, I am a new user of Open-MPI, and I need to use 2 kinds of programs on an unique cluster : 1) MPI based programs 2) Others, using TCP and UDP In order to get my non-MPI programs run, I need to know which ports may be used by MPI programs. Is there a way to know/set the range of the ports used by MPI programs ? Thanks Laurent.
Re: [OMPI users] users Digest, Vol 261, Issue 4
> -Original Message- > Date: Thu, 20 Apr 2006 19:35:27 -0400 > From: "Jeff Squyres \(jsquyres\)" > Subject: Re: [OMPI users] Open-MPI and TCP port range > To: "Open MPI Users" > Message-ID: > > Content-Type: text/plain; charset="us-ascii" > > > That being said, we are not opposed to putting port number controls in > Open MPI. Especially if it really is a problem for someone, not just a > hypothetical problem ;-). But such controls should not be added to > support firewalled operations, because -- at a minimum -- unless you do > a bunch of other firewall configuration, it will not be enough. This point is a real problem for me (but I may be the only one in the world...). I have to build a system that uses MPI softwares and non MPI COTS. I can't change TCP ports used by the COTS. Restricting MPI TCP/UDP port range loks like being the best solution for me.
Re: [OMPI users] Open-MPI and TCP port range
> -Original Message- > Date: Thu, 20 Apr 2006 19:35:27 -0400 > From: "Jeff Squyres \(jsquyres\)" > Subject: Re: [OMPI users] Open-MPI and TCP port range > To: "Open MPI Users" > Message-ID: > > Content-Type: text/plain; charset="us-ascii" > > > That being said, we are not opposed to putting port number controls in > Open MPI. Especially if it really is a problem for someone, not just a > hypothetical problem ;-). But such controls should not be added to > support firewalled operations, because -- at a minimum -- unless you do > a bunch of other firewall configuration, it will not be enough. This point is a real problem for me (but I may be the only one in the world...). I have to build a system that uses MPI softwares and non MPI COTS. I can't change TCP ports used by the COTS. Restricting MPI TCP/UDP port range loks like being the best solution for me.
[OMPI users] Checking the cluster status with MPI_Comm_spawn_multiple
Hi, Before starting programs on my cluster, I want to check on every CPU if it is up and able to run MPI applications. For this, I use a kind of 'ping' program that just send a message saying 'I'm OK' tu a superviser program. The 'ping' program is sent by the superviser on each CPU by the MPI_Comm_spawn_multiple command. It works fine when every CPU is up, but when one is down, my superviser stops when calling the MPI_Comm_spawn_multiple command. So the questions are : * 'What am I doing wrong ?' * 'Is there a other way to check my CPUs ?' Thanks for your help. Laurent.
[OMPI users] CPU use in MPI_recv
Hi,
I'm using Open-MPI 1.0.2 on a debian system.
I'm testing the MPI_recv function with a small C program (source code at the
end of the message). And I see that when I'm waiting a message, calling
MPI_recv, the CPU is used at 100 %.
Is that normal ?
Is there other ways to use a recv function (irecv, etc) that is not using the
CPU ?
Laurent.
Source code :
#include
#include
#include
int main(int argc, char *argv[])
{
int rc;
int numtasks, rank;
int myint = 0;
rc = MPI_Init(&argc, &argv);
if(rc != 0) {
printf("open error\n");
MPI_Abort(MPI_COMM_WORLD, rc);
}
MPI_Comm_size(MPI_COMM_WORLD, &numtasks);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
printf("from cpu_test : number of tasks : %d. My rank :%d\n", numtasks,
rank);
MPI_Recv(&myint, 1, MPI_INT, MPI_ANY_SOURCE, MPI_ANY_TAG,
MPI_COMM_WORLD, NULL);
printf("message received\n");
MPI_Finalize();
exit(0);
}
[OMPI users] mpi_comm_spawn_multiple and 'host' MPI_Info key
> Hi, > > I saw with great pleasure that in the last version of open mpi (1.1.1b4), the > 'host' MPI_Info key was available for the mpi_comm_spawn_multiple function. > > I tested it and I could spawn my processes on the wanted hosts. > Now, the question is : How can I tell mpi_comm_spawn_multiple which processor > to spawn, in a multi-processor architecture ? > > Maybe the 'host' MPI_Info key allows this feature, but I can't find the right > syntax. > > I also would like to know, if possible, when the v1.1.1 will be released. > > Thanks for your help, > Laurent. > >
[OMPI users] MPI_Comm_spawn_multiple and BProc
Hi,
I'm using MPI_Comm_spawn_multiple with Open MPI 1.1.1.
It used to work well, until I used the Bproc kernel patch.
When I use the Bproc patch, my program freezes when calling
MPI_Comm_spawn_multiple.
Does MPI_Comm_spawn_multiple and BProc can work together ?
Thanks,
Laurent Porez
Re: [OMPI users] MPI_Comm_spawn_multiple and BProc
Oups, sorry ! I followed these steps : 1) Install a debian system (sarge 3.1r2). 2) Use a 2.6.9 kernel patched with bproc 4.0.0pre8 (http://bproc.sourceforge.net) the options CONFIG_BPROC, CONFIG_ROMFS_FS, CONFIG_BLK_DEV_RAM, CONFIG_BLK_DEV_INITRD, CONFIG_TMPFS where activated via the 'make menuconfig' command. 3) install bproc 4.0.0.pre8 4) install beoboot-cm1.10 5) load bproc modules 6) Install open-mpi V1.1.1 Hope this will help. Thanks, Laurent. -- Could you please clarify - what "Bproc kernel patch" are you referring to? Thanks Ralph On 9/27/06 2:37 AM, "laurent.po...@fr.thalesgroup.com" wrote: > Hi, > > I'm using MPI_Comm_spawn_multiple with Open MPI 1.1.1. > It used to work well, until I used the Bproc kernel patch. > > When I use the Bproc patch, my program freezes when calling > MPI_Comm_spawn_multiple. > > Does MPI_Comm_spawn_multiple and BProc can work together ? > > Thanks, > Laurent Porez > ___
[OMPI users] OpenMPI 1.1.1 with Multiple Thread Support
Hi, Could you explain what's wrong with thread support ? Does it hang, or something else ? I'm developping an application using multiple processes with multiple threads, and I have to use MPI to make process communicate. Typically, I will have to use the following functions : - MPI_Comm_spawn_multiple, - MPI_Bsend, MPI_Recv, MPI_Irecv, MPI_Test - MPI_Barrier - MPI_Allgather Can this work with the actual version of Open-MPI (1.1.1) or a later one, or even with an other MPI library (free or commercial) ? Do I have to think about giving up MPI ? Thanks, Laurent.
[OMPI users] Error Handling Problem
Hi, I developped a launcher application : a MPI application (say main_exe) lauches 2 MPI applications (say exe1 and exe2), using MPI_Comm_spawn_multiple. Now, I'm looking at the behavior when an exe crashes. What I can see is the following : 1) when everybody is launched, I see the followings processes, using 'ps' : - the 'mpiexec -v -d -n 1 ./main_exe' command - the orted server used for 'main_exe' (say 'orted1') - main_exe - the orted server used for 'exe1' and 'exe2' (say 'orted2') - exe1 - exe2 2) I use kill -9 to 'crash' exe2 3) orted2 and exe1 finish. 4) with ps, I see it remains the following processes : mpiexec, 'orted1', main_exe 5) main_exe tries to send a message to exe1, using MPI_Bsend : main_exe gets killed by a SIG_PIPE signal So what I see is that when a part of an MPI application crashes, the whole application crashes ! Is there a way to get an other behavior ? For exemple, MPI_Bsend could return an error message. A few additionnal informations : - I work on linux, with Open-MPI 1.1.1. - I'm developping in C and C++. Thanks, Laurent.
Re: [OMPI users] Error Handling Problem
> From: George Bosilca > Subject: Re: [OMPI users] Error Handling Problem > To: Open MPI Users > Message-ID: > Content-Type: text/plain; charset=US-ASCII; delsp=yes; format=flowed > > How about changing the default error handler ? I did change the default error handler (using Mpi_Comm_set_errhandler) in the main_exe program. I replaced it with a printf. My error handler is never called, but main_exe receives a SIGPIPE signal. So the only solution I found is to catch SIGPIPE and forget it...> > It is not supposed to work, and if you find an MPI implementation > that support this approach please tell me. I know the paper > where you > read about this, but even with their MPI library this approach does > not work. which paper are you talking about ? > > Soon, Open MPI will be able to support this feature. Several fault > tolerant modes are under way, but no precise timeline yet. OK. I keep watching for new versions of Open MPI. Thanks, Laurent. > >Thanks, > george. > > On Oct 26, 2006, at 10:19 AM, laurent.po...@fr.thalesgroup.com wrote: > > > Hi, > > > > I developped a launcher application : > > a MPI application (say main_exe) lauches 2 MPI applications (say > > exe1 and exe2), using MPI_Comm_spawn_multiple. > > > > Now, I'm looking at the behavior when an exe crashes. > > > > What I can see is the following : > > 1) when everybody is launched, I see the followings processes, > > using 'ps' : > > - the 'mpiexec -v -d -n 1 ./main_exe' command > > - the orted server used for 'main_exe' (say 'orted1') > > - main_exe > > - the orted server used for 'exe1' and 'exe2' (say 'orted2') > > - exe1 > > - exe2 > > > > 2) I use kill -9 to 'crash' exe2 > > > > 3) orted2 and exe1 finish. > > > > 4) with ps, I see it remains the following processes : mpiexec, > > 'orted1', main_exe > > > > 5) main_exe tries to send a message to exe1, using MPI_Bsend : > > main_exe gets killed by a SIG_PIPE signal > > > > So what I see is that when a part of an MPI application crashes, > > the whole application crashes ! > > Is there a way to get an other behavior ? For exemple, MPI_Bsend > > could return an error message. > > > > A few additionnal informations : > > - I work on linux, with Open-MPI 1.1.1. > > - I'm developping in C and C++. > > > > Thanks, > > Laurent. >
[OMPI users] spawn on a cluster with 2 Ethernet interfaces
Hi, I have to spawn multiple slaves processes on a cluster, from a unique master process. The open mpi distribution I use is 1.1.2. I'm using a HP cluster, with 2 ethernet NICs on each machine. My problem was a freeze of master when calling mpi_call_spawn_multiple, and of slaves when calling MPI_Init. This appened when I tried to spawn on multiple hosts (worked well on a unique host). After working on the problem, I discovered that when I disabled eth1 on the hosts, everything got fine... The same behavior appens fortunately when I use the "--mca btl_tcp_if_include eth0" parameter. what is strange is that the problem stays if I use one of the followings : "--mca btl_tcp_if_include eth1" "--mca btl_tcp_if_exclude eth1" "--mca btl_tcp_if_exclude eth0" Is it impossible to use 2 Ethernet NICs at the same time for MPI applications ? Will I have to always use eth0, and not eth1 for MPI communications ? thanks, Laurent.
[OMPI users] Choosing the processor Id when spawning a process
Hi, I have to spawn a set of processes on multiple hosts, with my own mapping pattern, including processor ID, for example : * process 1 on cpu0 of host 1 * process 2 on cpu1 of host 1 * process 3 on cpu1 of host 1 * process 4 on cpu0 of host 2 * process 5 on cpu1 of host 2 I see that only the "host" MPI_Info parameter can be used (see ompi_comm_start_processes() ), but other kinds of mapping could be used : create_app( ), in orte\tools\orterun\orterun.c may handle ORTE_APP_CONTEXT_MAP_ARCH or ORTE_APP_CONTEXT_MAP_CN mapping, that would be perfect for me. Is there a next released that will take care of this, or is it of no use for most MPI users ? Thanks, Laurent.
[OMPI users] return from MPI_Comm_spawn
Hi, I see that when a master process spawns slave processes, MPI_Comm_spawn() does not return until all MPI_Init() function ends in slave processes. Is there a way to set a time-out, or something to detect when an error occurs in a slave process ?
Re: [OMPI users] How to set paffinity on a multi-cpu node?
I agree with this solution, for the machinefile. Using mpiexec or a spawn command, you can add the cpu number attached to the hostname : mpiexec -host [hostname]:[cpu number] -n 1 mpi_test or, for MPI_Comm_spawn : MPI_Info_set( mpi_info, "host", "[hostname]:[cpu number]" ); Cheers, Laurent. > > In the machinefile, add for each node with M cpus: > myhost@mydomain slots=N cpus_allowed=, > being the subset of 0..M-1 in some yours-to-decide format and > with yours-to-decide default values. > > Best Regards, > Alexander Shaposhnikov > > On Wednesday 29 November 2006 06:16, Jeff Squyres wrote: > > There is not, right now. However, this is mainly because > back when I > > implemented the processor affinity stuff in OMPI (well over a year > > ago), no one had any opinions on exactly what interface to expose to > > the use. :-) > > > > So right now there's only this lame control: > > > > http://www.open-mpi.org/faq/?category=tuning#using-paffinity > > > > I am not opposed to implementing more flexible processor affinity > > controls, but the Big Discussion over the past few months is exactly > > how to expose it to the end user. There have been several formats > > proposed (e.g., mpirun command line parameters, magic MPI > attributes, > > MCA parameters, etc.), but nothing that has been "good" and "right". > > So here's the time to chime in -- anyone have any opinions on this? > > > > On Nov 25, 2006, at 9:31 AM, shap...@isp.nsc.ru wrote: > > > Hello, > > > i cant figure out, is there a way with open-mpi to bind all > > > threads on a given node to a specified subset of CPUs. > > > For example, on a multi-socket multi-core machine, i want to use > > > only a single core on each CPU. > > > Thank You. > > > > > > Best Regards, > > > Alexander Shaposhnikov
