Jeff,
I issue lamboot in a driver that later on calls mpirun. I use the process id of the driver process to define LAM_MPI_SOCKET_SUFFIX so I shouldn't run into the big bang. Anyway, in the driver I will check if the job is running under PBS and modify the variable only if it's not. Does LAM detect an LSF queue in the same way?
Thanks,
Andras
Jeff Squyres <jsquyres_at_[hidden]> wrote:On Tue, 8 Jul 2003 sutovas_at_[hidden] wrote:
> From Jeff I know that the LAM comands lamboot/lamnodes/lamhalt recognize
> when jobs are run under PBS and they come up with a unique location for
> the socket. I presume that they effectively change the value of
> LAM_MPI_SOCKET_SUFFIX. Would it hurt if I also have this env variable
> defined or would LAM simply overwrite the value?
No, it will not hurt. If you set that variable, it will override the PBS
auto-detection stuff in LAM.
However, unless you can guarantee that users will never run more than one
PBS job on a node simultaneously, you should leave it dependant upon the
PBS job ID (or some other unique value) somehow. Otherwise, you can have
a "big bang" where multiple LAM universes collide. :-)
Note that the environment variable name has been deprecated in 7.0 --
although $LAM_MPI_SOCKET_SUFFIX works, we're encouraging everyone to
switch to $LAM_MPI_SESSION_SUFFIX (there's LAM_MPI_SESSION_PREFIX if you
need it, too).
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
_______________________________________________
This list is archived at http://www.lam-mpi.org/MailArchives/lam/
---------------------------------
Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!
|
