Issue:
TORQUE is only tracking resource usage of processes on the first node in the nodelist.
Symptom:
- MPICH 3.1.
- RHEL/CentOS 6.x
When using mpiexec to launch processes across multiple nodes, the processes are successfully launched on the correct nodes, but torque is only tracking resource usage of processes on the first node in the nodelist.
Solution:
Use (https://www.osc.edu/~djohnson/mpiexec)
Tags: mpiexec, not tracking, resources