Adding Hosts to the Virtual Machine

This section also applies to hosts started via a host file, because the same mechanism is used in both cases. The master pvmd starts up, reads the host file, then sends itself a request to add more hosts. The PVM console (or an application) can return an error when adding hosts to the virtual machine. Check the pvmd log file on the master host and the failing host for additional clues to what went wrong.

No such host - The master pvmd couldn't resolve the the host name (or name given in ip= option) to an IP address. Make sure you have the correct host name.

Can't start pvmd - This message means that the master pvmd failed to start the slave pvmd process. This can be caused by incorrect installation, network or permission problems. The master pvmd must be able to resolve the host name (get its IP address) and route packets to it. The pvmd executable and shell script to start it must be installed in the correct location. You must avoid printing anything in your .cshrc (or equivalent) script, because it will confuse the pvmd communication. If you must print something, either move it to your .login file or enclose it in a conditional:

    if ( { tty -s } && $?prompt ) then
        echo terminal type is $TERM
        stty erase '^?' kill '^u' intr '^c' echo
    endif

To test all the above, try running the following command by hand on the master host: rsh host $PVM_ROOT/lib/pvmd -s

where host is the name of the slave host you want to test. You should see a message similar to the following from the slave pvmd and nothing else:

    [pvmd pid12360] slave_config: bad args
    [pvmd pid12360] pvmbailout(0)

Version mismatch - This message indicates that the protocol revisions of the master and slave pvmd are incompatible. You must install the same (or compatible) versions everywhere.

Duplicate host - This message means that PVM thinks there is another pvmd (owned by the same user) already running on the host. If you're not already using the host in the current virtual machine or a different one, the socket address file (§) must be left over from a previous run. Find and delete it.