LAM/MPI logo

LAM/MPI General User's Mailing List Archives

  |   Home   |   Download   |   Documentation   |   FAQ   |   all just in this list

From: Faisal Iqbal (faisal749_at_[hidden])
Date: 2007-02-06 01:10:03


Dear Tim,
 
 I tried your solution but unfortunately it didn't work out completely.
 
 When trying mpirun n0 /common/hello it worked and gave correct output. [n0 is my head node]
 
 However when trying to expand it to 2 pcs, i got the same error [It seems that at least one of the processes didn't invoke MPI_INIT.....] Here is the command i tried
 mpirun n0-1 /common/hello
 mpirun n0,1 /common/hello
 
 Below is the output for this
 -----------------------------------------------------------------------------
 It seems that [at least] one of the processes that was started with
 mpirun did not invoke MPI_INIT before quitting (it is possible that
 more than one process did not invoke MPI_INIT -- mpirun was only
 notified of the first one, which was on node n0).
 
 mpirun can *only* be used with MPI programs (i.e., programs that
 invoke MPI_INIT and MPI_FINALIZE). You can use the "lamexec" program
 to run non-MPI programs over the lambooted nodes.
 -----------------------------------------------------------------------------
 
 And trying -v with mpiruin gave following output
 
 
 3227 /common/hello running on n0 (o)
 2792 /common/hello running on n1
 -----------------------------------------------------------------------------
 It seems that [at least] one of the processes that was started with
 mpirun did not invoke MPI_INIT before quitting (it is possible that
 more than one process did not invoke MPI_INIT -- mpirun was only
 notified of the first one, which was on node n0).
 
 mpirun can *only* be used with MPI programs (i.e., programs that
 invoke MPI_INIT and MPI_FINALIZE). You can use the "lamexec" program
 to run non-MPI programs over the lambooted nodes.
 -----------------------------------------------------------------------------
 
 And below is the output for ldd /common/hello
 
         linux-gate.so.1 => (0x00879000)
         libutil.so.1 => /lib/libutil.so.1 (0x009c6000)
         libdl.so.2 => /lib/libdl.so.2 (0x005ac000)
         libstdc++.so.6 => /usr/lib/libstdc++.so.6 (0x00111000)
         libm.so.6 => /lib/libm.so.6 (0x00586000)
         libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x007c4000)
         libpthread.so.0 => /lib/libpthread.so.0 (0x006cf000)
         libc.so.6 => /lib/libc.so.6 (0x0045a000)
         /lib/ld-linux.so.2 (0x0043c000)
 
 I can't get the reason for it working on one node and not on anyother :(
 
 Best Regards,
 Faisal
 
Tim Prins <tprins_at_[hidden]> wrote: Faisal,

>From the information you've given I can't be a whole lot of help, but I would
suggest the following to start figuring out the problem:
1. Try running lamboot and running your program on a single node [something
like mpirun -n 1 /common/hello]
2. If this works, try again on multiple nodes

If any of these fail, please add '-v' to the mpirun command line and send the
output.

Also, are you using the provided mpicc to compile the program? What is the
output of 'ldd /common/hello'?

Tim

On Wednesday 31 January 2007 12:50 am, Faisal Iqbal wrote:
> Dear ALL
> I've started a university project that requires me to build a cluster and
> solve a problem on it. I've setup cluster. I tried lamboot and it worked
> perfectly. Then i created a simple "Hello World" program [which i copied
> from LAM User Guide] and it ciompiled successfully. However when running
> this program [mpirun C /common/hello] where /common is nfs share and
> mounted on all workstations while hello is the executable, i get the
> following error
>
>
> ---------------------------------------------------------------------------
>-- It seems that [at least] one of the processes that was started with
> mpirun did not invoke MPI_INIT before quitting (it is possible that more
> than one process did not invoke MPI_INIT -- mpirun was only notified of the
> first one, which was on node n0).
>
> mpirun can *only* be used with MPI programs (i.e., programs that
> invoke MPI_INIT and MPI_FINALIZE). You can use the "lamexec" program
> to run non-MPI programs over the lambooted nodes.
>
> ---------------------------------------------------------------------------
>--
>
> I searched a lot both on google and this mailing list but couldn't find
> anything applicable. I'll appreciate if any of you may help me out from
> this :)
>
> Best Regards,
> Faisal
>
>
> ---------------------------------
> TV dinner still cooling?
> Check out "Tonight's Picks" on Yahoo! TV.

_______________________________________________
This list is archived at http://www.lam-mpi.org/MailArchives/lam/

 
---------------------------------
It's here! Your new message!
Get new email alerts with the free Yahoo! Toolbar.