Sry its Rocks Cluster 6.1.1
On my spare time I am playing with hash cracking while taking college courses for Network Security.
The Project:- 7 X Dell PE 2950 (2) QC / 4GB RAM / 73GB HD ( nodes )
- 1 X Dell PE 2900 (2) DC / 4GB RAM / 3x73GB RAID 0 ( master )
- 1 X gigabit switch
- Downloaded john-1.7.9-jumbo
- Downloaded short & long password lists 250MB & 4.5GB
Whats been done:- I have setup the master node & included and installed the other slave nodes via "insert-ether" with no real issue.
- Compute nodes register on the cluster and are detailed in ganglia
- Created a general user account on the system
- As normal user edited john's make file for mpi and ran "make linux-x86-64" as per john's mpi instructions.
- Created machines file in normal user path for involved compute node
- Made the standard HPL.dat file in normal user path
The Problem:
I can run john --test locally on master node without issue "mpirun -np 16 ./john --test"
But when I run the same test using the machine file I get errors that the directory or command is not found on remote nodes. "mpirun -np 16 -machinefile machinefile ./john --test". there is a /exports/home/"normal user" DIR on the master node -but- there is no related files including john's dir on the nodes in relation to where it should be according to ROCK documentation. I have resynced the NFS but no change.
The Question:
Am I totally missing something in the case of my user programs and files on the master node and how it is propagated to the compute nodes?
Or am I being dumb and totally not setting up MPI calls correctly?
Other than the Rocks Cluster initial setup is smooth...
Any idea's are appreciated
Ty
