On Apr 15, 2005, at 8:56 PM, Understudy wrote:
> Okay, Maybe one of things that is making this so hard for me to
> assimilate is my preconceived notion of how a cluster works. My
> understanding was that there was a master computer and the rest were
> slaves. Even though they functioned as one. The master would be the
> one the user interfaced with.
Yes and no. It depends on how you setup your cluster; many do have a
"master" or "head node", but it's a local policy decision (e.g., if you
run NFS, it is typically your NFS server; if you run a batch scheduler,
it's typically the "main" scheduler node, ....etc.). LAM does not care
about such things -- it'll run on whatever machines you tell it to run
(including, for example, a laptop).
This isn't really the right list for cluster-setup kinds of questions;
you might want to head over to the beowulf list for such kinds of
things. There's also oodles of web pages out there about how to setup
clusters; you might want to google around a bit.
> The Makefile in FreeBSD currently shows this:
> # New ports collection makefile for: lam
[snipped]
> RUN_DEPENDS+=
> ${SITE_PERL}/File/Temp.pm:${PORTSDIR}/devel/p5-File-Temp
> .endif
> There is more to the Makefile but I don't think it was needed.
Not knowing anything about ports, that Makefile looked ok to me.
> I guess I could install it manually but it defeats the point of having
> a port or package.
Per the advice from the other post about installing from ports, this
may be reasonable. I really have no basis for recommendations here.
> It still leave me with a lot of unanswered questions. The main one
> being that since I have installed it as a port should I deinstall it
> or can I make it work by correcting a few things with the existing
> installation.?
What I've been trying to say is that I'm not the right guy to answer
this -- my team has nothing to do with, and no knowledge of this
"port", nor how it installed or setup. A seg fault can happen for a
million reasons; you really haven't provided any additional specific
information other than "it doesn't work," making it somewhat difficult
for us to try to help.
A logical first step could be to try to install LAM via tarball and see
if the same seg fault occurs. If it does, at least that narrows down
the the problem: the "port" install is probably ok, and there is likely
a problem with your local configuration or LAM itself (potentially LAM,
because laminfo should not be seg faulting). If it doesn't, then
perhaps the "port" installation is somehow to blame. Can you try this
and see what happens?
LAM *should* work on all POSIX-like systems, but as is plainly obvious
by this point, we don't have any FreeBSD boxen to test with, so perhaps
there's a portability quirk there that we're not handling correctly.
If so, we're certainly interested in fixing it.
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
