Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

Re: Firefly runs on 1 node on cluster and doesn't run on 2 nodes.


Problem was in installation-specific settings of Cleo. In monopolistic rate (without using Cleo) Firefly runs on 2 nodes successfully.


On Mon Sep 27 '10 1:21pm, Olga wrote
>I try to work with Firefly on cluster, witch has 10 nodes and 8 cores on each node. On it SUSE Linux Enterprise Server 10, Cleo manager and Intel MPI are installed.
>I download Firefly version 7.1.G download links, Linux Intel MPI, dynamically linked NPTL-based version.
>My command line is:
>mpirun -np 16 /home/kremov/ff/firefly -r -f -p -stdext -i /home/kremov/ff/1/bench01.inp -o /home/kremov/ff/1/bench01.out -ex /home/kremov/ff -t /home/kremov/tmp

>If I write -np 8, then bench01.out and bench01.dat appear in folder "1", program works normally. But if -np 16, then in "1" nothing appears, and in cluster report file this error is written:

>MPI process terminated unexpectedly
>Exit code -5 signaled from node-01
>cleanupKilling remote processes...DONE

>Why does it happens?

>If -np 8, then in bench01.out there is such lines:
> max    # of   cores/package :   4
> max    # of threads/package :   4
> max     cache sharing level :   2
>What this means?

>Thanks in advise!

[ Previous ] [ Next ] [ Index ]           Wed Sep 29 '10 9:27pm
[ Reply ] [ Edit ] [ Delete ]           This message read 683 times