Do I really have to run makelocalrc on each node?

What’s the purpose of bin/localrc.{machinename} file and the local lib directories that it wants to create on each node?

Hi Pam,

For a network install, the installer must assume your systems are heterogeneous hence each do need there own local configuration.

If your cluster is homogeneous, then it might be easier to install the compilers using the ‘single system install’ option on a shared file system. In this case only a single localrc will be created.

Hope this helps,
Mat

Mat, thanks for the reply.

We have a couple different clusters, with a mix of RHEL4 and RHEL5 OS.
So you’re saying that in a situation like this, the compiler needs to know about the specifics of the RHEL4 systems and the RHEL5 systems?
The RHEL4 systems do have a different (older) glibc version and gcc version (this is dictated by Red Hat) as compared to the RHEL5 systems.

I was thinking about duplicating the localrc.{machinename} file for each of the RHEL5 machines, then running makelocalrc on one of the RHEL4 machines and likewise duplicating the result.

One last thing, some of the RHEL4 nodes have Opteron processors, and some have Xeon. Will this make a difference in the localrc files or in the per-machine libraries that get installed by makelocalrc?

thanks!