Newsgroups: comp.parallel.pvm
From: link@enif.astro.indiana.edu (Robert Link)
Subject: Re: arcane error message when starting pvm
Organization: Indiana University
Date: 16 Aug 1996 19:14:54 GMT
Message-ID: <4v2hbe$gjd@usenet.ucs.indiana.edu>


wjiang@apollo.hp.com (Weicheng Jiang) wrote:
>
>Ok, I dug up a couple of replies from Bob Manchek regarding this error,
>hope they'd help you find the cause. Looks like rsh might be the culprit
>here, what machines are you running on?
>

We are running on RS6000's.  I checked the RS6K.def file, and it does
not have the fix below, but, unfortunately, anything involving
rebuilding PVM means convincing the sysadmins to do it, and they are
skeptical. 

>
>|> |> I am able to start pvmd from a SUNOS machine ... and add Solaris hosts.
>|> |> I am unable to start pvmd from a Solaris machine ...  I get
>|> |>
>|> |> [t80040000] ready  3.3.9   Mon Sep 25 16:16:03 1995
>|> |> [t80000000] pl_startup() pvmd@cbi4: EOF
>|> |> [t80040000] startack() host cbi4 expected version
>|> |> [t80000000] pl_startup() pvmd@cbi4: EOF
>|> |> [t80040000] startack() host cbi4 expected version
>|>
>|> [but there doesn't seem to be anything wrong]
>|>
>|> If anyone else is having this problem, it may be because PVM is trying
>|> to use /usr/ucb/rsh and you don't have the sun compatability package
>|> installed.  Edit $PVM_ROOT/conf/SUN4SOL2.def and add
>|>     -DRSHCOMMAND=\"/bin/rsh\"
>|>
>|> to the ARCHCFLAGS line.  Then rebuild PVM (at least src/ARCH/startup.o).
>|>


>
>> more pvml.2028
>> [t80040000] ready  3.3.7   Wed Jan 17 11:59:03 1996
>> [t80000000] pl_startup() pvmd@btrtx1: EOF
>> [t80040000] startack() host btrtx1 expected version
>> [t80000000] pl_startup() pvmd@btrtx6: EOF
>> [t80040000] startack() host btrtx6 expected version
>>
>> What's wrong in my configuration.
>
>Hi,
>
>What happens when you run the slave pvmd command by hand:
>
>    rsh btrtx1 '$PVM_ROOT/lib/pvmd -s'
>
>You should get:
>
>    slave_config: bad args
>    pvmbailout(0)
>If you see something else, there will be problems.
>

It is worth noting that this is, indeed the result I get when I try
this experiment.  Also worth noting is that I I get the same set of
errors (i.e. the EOF and the ``expected version'') when I set
everything to manual startup (* so=ms), which, unless I am horribly
mistaken, should not depend in any way on rsh.

-- 
-r 	| Some people never find it; 
	| Some only pretend, but me,
        | I just want to live happily ever after
        | Every now and then

