Need assistance on setting up MPICH cluster

Hello everyone,

I am ruing into a few issues with the SSH and MPICH executing. From some previous questions that I asked, I was able to progress to a point where I executed the mpi_hello.c program.

For reference, I am working on following this tutorial on setting up MPICH: https://help.ubuntu.com/community/MpichCluster

I created a directory in root called clusterFiles and I created a user on all of the nodes called clusterUser (clusteruser). I exported clusterFiles and I mounted clusterFiles in all of the nodes. Also, I changed ownership of clusterFiles to clusterUser on the master node. I also changed the home directory of clusterUser to be /clusterFiles.

I created an ssh key for clusterUser on the master node and I added the key to the authorized lists. I installed a keychain on the all nodes and on the master node I edited the .bashrc as specificed in the guide (I copied what was in the guide into .bashrc)

I also installed MPICH2 and GCC on all nodes.

I edited the machine file for my scpecific cluster.

However, when I go to execute the MPI hello_world.c program, this is where the errors occur.

I copied and pasted the code on the guide into a .c file and called it mpi_hello.c (This was done on the master node).

In the guide, the last part he just calls `mpicc [arguments]` and `mpiexec [arguments]`. However, when I go to call mpicc, I need to `sudo mpicc [arguments]`. Is this a problem I should be conceed with or would this be the proper way that it should it be done?

When I run mpiexec (without sudo), I recieve that following errors:

clusteruser@rgcluster2blade1:~$ mpiexec -n 7 -f machinefile ./mpi_hello

[mpiexec@rgcluster2blade1] HYDU_parse_hostfile (./utils/args/args.c:323): unable to open host file: machinefile

[mpiexec@rgcluster2blade1] mfile_fn (./ui/mpich/utils.c:341): error parsing hostfile

[mpiexec@rgcluster2blade1] match_arg (./utils/args/args.c:153): match handler retued error

[mpiexec@rgcluster2blade1] HYDU_parse_array (./utils/args/args.c:175): argument matching retued error

[mpiexec@rgcluster2blade1] parse_args (./ui/mpich/utils.c:1609): error parsing input array

[mpiexec@rgcluster2blade1] HYD_uii_mpx_get_parameters (./ui/mpich/utils.c:1660): unable to parse user arguments

[mpiexec@rgcluster2blade1] main (./ui/mpich/mpiexec.c:153): error parsing parameters

Are these files something that forgot to install? At first, I am thinking that I need sudo in front of mpiexec. So when I perform: `sudo mpiexec [arguments]` it "runs" but coects to the SSH cluster as root when I need it to coect as clusteruser.

My main conce is that he is not executing his commands as root. I am wondering if there is a step that is implied or at least there is a command that I was suppose to execute but didn't?

Also, I noticed that when I tried changing ownership of clusterFiles to clusterUser on the other nodes, I would get an operating not permitted error (I was root when I did this command). My thinking is that since I changed the ownership on the master node, it propagated to the other nodes since they have same useame. So I was effectively changing the ownership to itself. Is this a correct thinking or is there more to it then that?

- - , .
.