[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
[ptp-dev] SDM scalability issue
|
Hi,
I am having a weird problem while debugging with PTP 2.1.0.
I have a cluster of 10 Power machines running Linux Fedora 9 and Open
MPI 1.3. I can run MPI applications from command line and from within
PTP on this cluster without problems, even when I use 10 processes
distributed on the 10 nodes I have.
However, when I try to debug the application (I am using the "Open MPI
Pi C Project" that comes with PLDT) I am able to debug without problems
using any number of processes on my cluster from 1 to 7. However, when I
switch to 8 processes or more, the debug session is started and
connected, SDM is started, but the processes don't suspend execution in
the beginning of the debug session. After a while, some processes may
get suspended, but in all cases one of two things happens: none of the
process get suspended or some of the processes (usually two of them)
have their debug session broken and the debug session enters an
inconsistent state from which I can only get out when I cancel the debug
completely.
I've already tried to launch the 8 processes using different
combinations of the machines in the cluster. I always get the same
problem. On the other hand, if I try different combinations of machines
with 7 processes, it always works fine.
Does anyone already faced something similar to that?
Best regards,
Leonardo Garcia