README.md

MPAS

Instructions for running diverge after Randall's instructions at Launch a starcluster named

$ starcluster start -c <cluster> <cluster>
$ starcluster sshmaster <cluster>

Now you are on the cluster on AWS. Next is to clone the mpasdocker git repository It is very important that you have to clone the mapsdocker git repository to the local directory where the shared data volume is mounted.

$ cd /home/mpas
$ git clone https://github.com/guanxyz/MPAS
$ cd MPAS

Edit hostfile to represent the number of nodes on your <cluster>. In this case, we use a two-nodes cluster Edit machinefile to represent the number of nodes and processors on your <cluster>

This builds your docker, which is one process per node.

$ mpirun -hostfile hostfile --mca btl_tcp_if_include eth0 ./mpibuild_mpas.sh --verbose --output-filename=mpibuild

This runs your simulation in docker, which is n processes per node.

$ mpirun -hostfile hostfile --mca btl_tcp_if_include eth0 ./mpirun_mpas.sh --verbose --output-filename=mpirun

Description of Files

* Dockerfile -- build of MPAS and I/O libraries into a Docker container * mpibuild_mpas.sh -- script that builds the docker container on each node * mpirun_mpas.sh -- script to launch H3D with docker containers on each node * runmpas.sh -- script that runs H3D inside a docker container on the master node. DO NOT call this directly. It is called by mpirun_mpas.sh