If you want to run Totalview for a parallel job using several nodes, you might need following:
Currently, Totalview in PPPL cluster sets default accessing shell as “rsh”. So if you try to run Totalview, the program will try to use “rsh”, which is not supported in PPL cluster, and you’ll meet error message like ” connection refused ” with several lines showing trials of access to other nodes using rsh.
To change the default value, you might need to add a following line to your “.cshrc” (Cshell) or “.bashrc”(bash) as
setenv TVDSVRLAUNCHCMD ssh (cshell)
export TVDSVRLAUNCHCMD=ssh (bash)
In addition, it’ll be necessary to use interactive job(“-I” option) to run totalview. Note that you need one more option “-V” to transfer your environmental variable(esp. for DISPLAY) to allocated nodes. If not, you’ll meet X11 forwarding problem, and an error message telling environmental variable DISPLAY is not properly set.
After you get interactive nodes, you can run totalview(“-tv” option) with MPICH as
mpirun -tv -np (# of CPUs) (executable file)
