Solution

If you are having difficulty getting PVM running correctly on your system, in the first instance please check the following:

1. Check that the environment variable

$PVM_ROOT

is set correctly and globally on all machines within the PVM cluster. Normally this should be $GOLD_DIR/pvm3 where $GOLD_DIR would be <INSTALL_DIR>/GoldSuite_2016/GOLD

2. Check that your system temporary area is not full. We have occasionally heard of cases where PVM could not start correctly because /tmp on the user's machine was full.

Once you have performed these checks, you can begin diagnosing the root cause of the problem. The UNIX GOLD distribution includes a PVM diagnostics script called "test_pvm.sh". To run this script, please execute the command:


$GOLD_DIR/bin/test_pvm.sh

and follow the on-screen instructions. If you are unable to interpret the information generated by this script, please send the entire output by email to support@ccdc.cam.ac.uk and we will diagnose any PVM problems you may have.

Additional diagnostic information can be obtained from various files that can be found on the machines within your PVM cluster. In particular, the PVM log files are often very useful. They take the form:

/tmp/pvml.<user id>

and are generated on both the PVM master machine and the PVM slave machines. They can contain relevant information (or sometimes lack expected lines) that indicates the source of the problem.

For example, if PVM is configured correctly you should expect to see the text line

Running on <platform type>

somewhere in the PVM log file on the PVM master machine.

For further information, please consult the troubleshooting guide at http://www.netlib.org/pvm3/book/node1.html

Please note that the 2019 CSD Release is the last to include the ability to use GOLD with Parallel Virtual Machine (PVM) for UNIX-based platforms. We recommend using a grid-engine based environment rather than PVM.


« Return to search results