On Jan 26, 2005, at 7:36 PM, Dale Harris wrote:
> This was something I posted to the bproc mailing list already and I
> figured I should post it here, too. The thing is I'm trying to get
> lam-7.1.1 to work the bjs, which is the bundled scheduler for bproc.
>
> So anyone tried to get this too work? Doesn't appear lamboot is very
> happy with bjssub. Any hints on how to get this to work?
We (the LAM developers) have not tried using BJS. I've just tried your
script, which I would have expected to work. It looks like the problem
is that the LAM daemons on the compute nodes are dying almost
immediately after finishing the handshake with lamboot. I'm not
exactly sure what BJS does that puts the lamd into such a funk, and
debugging is going a little slowly (on a good day, lamd debugging is
difficult. When you just barely have a useable gdb like when on a
BProc compute node, it's not a good day).
Anyway, I wanted to let you know that we are looking at this problem
and hope to have a fix / workaround shortly.
Brian
--
Brian Barrett
LAM/MPI developer and all around nice guy
Have a LAM/MPI day: http://www.lam-mpi.org/
|
