On Jan 13, 2006, at 7:02 AM, zayar win wrote:
> I have a problem with LAM/MPI. I mounted head’s /home and /usr/
> local in every client’s /home and /usr/local directory via nfs. I
> intalled LAM 7.1.1 only head’s /usr/local directory. When I run my
> C program error occurred, shown in below.
>
> a.out: /lib/i686/libc.so.6: version ‘GLIBC_2.3.2’ not found
> (required by a.out)
This looks like LAM/MPI was compiled on one machine and then
attempted to be run on another where the versions of Glibc did not
match. You'll likely have the same problem with *any* C program that
was compiled on machine A and then tried to run on machine B.
Perhaps the version and/or distro of Linux is different between your
head node and your compute nodes? (if this is the case, my personal
recommendation for a simple beowulf-style cluster is to make them all
the same -- having a head node that is not able of producing
executables that can be run on the compute nodes can be a real
logistical battle)
See LAM's definition of "heterogeneous" in the "Heterogeneity and LAM/
MPI" section of the FAQ:
http://www.lam-mpi.org/faq/
--
{+} Jeff Squyres
{+} The Open MPI Project
{+} http://www.open-mpi.org/
|
