I have a problem with a particular software pacakage. Unfortunately
the vendor doesn't support "Network Access" for Linux.
I have a Gentoo/OpenMosix(2.4.22-1) cluster that is used to run
compute intense modeling jobs. The jobs run fine when run locally
(started within a shell), but fails (seg fault) when called via ssh
(or rsh) from another machine.
I narrowed it down to the c5.0 software, which itself runs fine
locally but fails when called remotely. I can't find anything else on
the machine that fails in this manner.
Also another Gentoo/OpenMosix (2.4.19-7) cluster which has the same
software on it works fine. I tried using the 2.4.19 kernel for my
problem cluster with the same results. Environment doesn't seem to be
an issue, because locally I stripped down the environment to next to
nothing and the program ran. I compared envs between the remote call
and local and there wasn't anything different.
In a nutshell:
Does anyone have any ideas what would cause a seg fault using:
machine1 % ssh machine2 "/usr/local/c5.0/bin/c5.0"
But the following works:
machine1 % ssh machine2
machine2 % /usr/local/c5.0/bin/c5.0
TIA,
Paul
|