i need help... badly
William Gropp
gropp at mcs.anl.gov
Wed Sep 18 14:53:47 PDT 2002
At 07:39 PM 9/17/2002 -0700, Mary Francis Moll wrote:
>I ran the tstmachines script but errors occurred.
>Running the true command on each machine went fine but
>when the ls command was run, I get the error that
>~/bin/mpichfoo was not s file or directory. And this
>only occurred for the other machines. At the end of
>the error message, the only machine that seems to be
>available was the machines I ran the script on.
>
>Please, please, please, I need help...
This means that the machines in your cluster do not have a shared file
system, for example, they don't use NFS cross mounts to make a common file
system visible (with the same path name) to each machine in your
cluster. The mpirun script relies on having a shared file system, though
it is possible to run MPICH programs without one. If you can't setup a
shared file system, you will need to arrange to copy the executables to
each machine first, and then use a "procgroup file" (covered in the
documentation) to specify which executables to run on each machine.
Bill
More information about the Beowulf
mailing list