LAM/MPI logo

LAM/MPI General User's Mailing List Archives

  |   Home   |   Download   |   Documentation   |   FAQ   |   all just in this list

From: Jeff Squyres (jsquyres_at_[hidden])
Date: 2004-09-15 04:42:56


That's an interesting trick!

If that works, that must mean that you have UDP connectivity between
all your nodes. Is there a reason that you're using globus, then? Are
they different authentication relams, perhaps?

On Sep 10, 2004, at 4:48 PM, redirecting decoy wrote:

> Hmm, I was afraid the I would have to install globus
> on all of the nodes. I wanted to avoid doing this,
> because configuring globus is a bit troublesome.
> Attempting to do this correctly across all my nodes is
> a headache waiting to happen. I did find another
> solution that seems to work, although it's not very
> elegant.
>
> This is what I did:
> "lamboot -v -ssi boot globus machines.globus" one of
> the frontend nodes. This creates the globus
> connection between my two frontends. Then I ssh'ed
> into each frontend, and did a:
> "lamgrow -v compute-0-X" X=Number of the node 0-3
> doing this on both frontends, seems to create a lam
> universe fine. Obviously the above command can be put
> into a script. Anyway, this seems to work just fine,
> but I am still testing to see if any errors/problems
> arise. Although, I would like a better way of doing
> it. I am open to any suggestions. Perhaps an elegant
> solution exists.
>
> If anyone has any ideas, please let me know.
>
> Thanks,
>
> -RD
>
>
> --- Vishal Sahay <vsahay_at_[hidden]> wrote:
>
>> Currenly LAM's support for Globus is only on
>> machines which have the
>> Globus gatekeeper running. It looks like your front
>> end machines are the
>> only two that have the gatekeepers and then they
>> kind of schedule jobs to
>> the internal nodes. LAM currently does not support
>> such configurations.
>> You need to have Globus installed and the gatekeeper
>> running on each
>> machine you wish to run LAM jobs on.
>>
>> -Vishal
>>
>>
>> On Fri, 10 Sep 2004, redirecting decoy wrote:
>>
>> # Hello everyone,
>> #
>> # I am using lam-mpi 7.0.6 and the Globus Tool Kit
>> 3.2.1
>> # on 2 different clusters with 4 nodes each. There
>> is a
>> # total of 10 machines, 2 Frontends and 8 Nodes.
>> These
>> # are ROCKS based. I have been able to get the Lam
>> # universe to boot using globus. However, I only
>> have
>> # the 2 frontends(grids) in the machine file. My
>> # question is, how do I get Lam to use all 10
>> machines,
>> # both frontends with the nodes behind them.
>> #
>> # Here is some more information on my setup:
>> # Each frontend has 2 nics, with the compute nodes
>> on 10
>> # dots. Globus is installed on the 2 Frontends, and
>> lam
>> # is installed on all of the machines.
>> #
>> # Frontend0.public: 192.168.10.16
>> # compute-0-0: 10.255.255.254
>> # compute-0-1: 10.255.255.253
>> # compute-0-2: 10.255.255.252
>> # compute-0-3: 10.255.255.251
>> #
>> # Frontend1.public: 192.168.10.17
>> # compute-0-0: 10.1.2.254
>> # compute-0-1: 10.1.2.253
>> # compute-0-2: 10.1.2.252
>> # compute-0-3: 10.1.2.251
>> #
>> # I created a file called machines.globus:
>> # machines.globus:
>> # Frontend0.public prefix=/usr/local/lam
>> # Frontend1.public prefix=/usr/local/lam
>> #
>> #
>> # And used the following command to boot.:
>> # lamboot -v -ssi boot globus machines.globus
>> #
>> # So this works fine, but how do I include the rest
>> of
>> # the nodes ? Any help will be greatly appreciated.
>> #
>> # Thanks in advance,
>> #
>> # -RD
>> #
>> #
>> #
>> # _______________________________
>> # Do you Yahoo!?
>> # Shop for Back-to-School deals on Yahoo! Shopping.
>> # http://shopping.yahoo.com/backtoschool
>> # _______________________________________________
>> # This list is archived at
>> http://www.lam-mpi.org/MailArchives/lam/
>> #
>> _______________________________________________
>> This list is archived at
>> http://www.lam-mpi.org/MailArchives/lam/
>>
>
>
>
>
> _______________________________
> Do you Yahoo!?
> Shop for Back-to-School deals on Yahoo! Shopping.
> http://shopping.yahoo.com/backtoschool
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>

-- 
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/