[GE users] Changing qmaster name

Thomas P Foley tfoley at uab.edu
Mon Apr 17 01:29:15 BST 2006

I am a novice SGE administrator as well as a novice cluster
administrator and have a problem that I am hoping this listserv can help
with. I recently installed a Linux high performance computing cluster
with Platform Rocks 4.0 which includes SGE 6.0u4. The problem that I
have is that for some reason the Platform Rocks installation chose the
wrong interface name to configure SGE on. For example the front end
server has three different network cards, one for Internet, one for a
compute network and one for a management network. The
/opt/gridengine/default/common/act_qmaster has the name
coosa.eng.uab.edu instead of the correct name of coosa.local. Therefore
when the compute nodes are installed they are looking for the qmaster on
a network they cannot talk to. What I need to figure out how to do is to
change the configuration of SGE to point to coosa.local. After reading
the documentation it would almost appear that I could stop all SGE
services on the front end node, edit act_qmaster and restart the
services. However the documentation states to use the name that is found
by running /opt/gridengine/utilbin/lx26-amd64/gethostbyname. When I do
this of course the name returns exactly coosa.eng.uab.edu. How can I
change the qmaster name so that the front end node as well as all the
compute nodes will point to the correct qmaster?

Thanks in advance for all your help and advise.

Tommy Foley
Director of IT
UAB School of Engineering
tfoley at uab.edu

More information about the gridengine-users mailing list