Thx alot Stuart for the reply it is so helpful , something else @ http://www.globusconsortium.org/tutorial/ch7/page_2.php there is a command to test the pbs and globus connection which is this command " globusrun-ws -submit -F https://nodeb.ps.univa.com:8443/wsrf/services/ManagedJobFactoryService -Ft PBS -c /bin/sleep 120 " but when i do test it it is not working and showing me services unavailable, is there any other way to test the PBS and globus ?
Thank you regards, sherif On Thu, Feb 26, 2009 at 5:55 PM, Stuart Martin <[email protected]> wrote: > > On Feb 25, 2009, at Feb 25, 11:50 AM, Sherif Nagy wrote: > >> Hello, >> >> I am a little bit new into grid computing and globus, I have been >> reading @ http://globus.org/toolkit/docs/4.2/4.2.1/admin/install/ and >> @ http://www.globusconsortium.org/tutorial/ for few days now. but i am >> a little bit confused , as my setup now having a job batch and >> scheduling system that is already working " Torque/PBS with Maui " one >> machine running pbs_server " which will be running globus as far as I >> understand " and 4 machines running the pbs mini server pbs_mom which >> connects to the PBS and everything works perfect, so if i understand >> correctly I'll install globus toolkit into the PBS server with >> --enable-wmfs-pbs flag > > Yes. But it is "--enable-wsgram-pbs" > >> and no need to install anything on the compute >> nodes "the other 4 machines" ? > > Right. > >> not even globus-ftp gridftp services >> for file management ? > > Well, GRAM assumes there is a shared file system for the compute cluster > (head node and worker nodes). Typically, there is a gridftp server running > on the same host as the gram service and used for file transfers from/to > remote clients to/from the head node. But you can configure the gridftp > server to run on a separate host and configure ws gram to use it. If there > are any file system path differences between these 2 hosts, then add those > filesystem mapping file. > > http://www.globus.org/toolkit/docs/4.0/admin/docbook/ch11.html#s-wsgram-Interface_Config_Frag-filesysmap > >> the globus will just submit the job to the pbs >> and the pbs will handle everything else ? > > With a shared file system, then PBS will schedule and execute the jobs. > Then gram will perform and remote file transfers before and after the job, > e.g. file stage in and out (and cleanup). > >> also if i have another >> cluster i'll have to install globus on the master node and link both >> globus setup together right ? > > You would do the same installation on the 2nd cluster. And now you can > submit gram jobs to each. > > Then there are client workload tools that will facilitate the execution to > multiple gram installations. > > Some populate tools for this are: Condor-G, GridWay, Swift, GEMLCA, AHE > > http://dev.globus.org/wiki/Incubator/Swift > http://www.gridway.org > http://www.cs.wisc.edu/condor/condorg/ > http://wiki.realitygrid.org/wiki/AHE_GENIUS > http://dev.globus.org/wiki/Incubator/GEMLCA > >> and what about the clients do they need >> to deploy globus also ? please advice and if there is any extra links >> will help me understand the big picture will be appreciated. > > The GRAM documentation is here: > http://www-unix.globus.org/toolkit/docs/4.0/execution/wsgram/ > >> >> Thank you >> regards, >> Sherif > >
