I’ll start with a summary of what I’d like to do. I have just read the basics of the terminology, so forgive me if I say something wrong.
I have an x86 host that I would like to use as a PBS server (including scheduler, communication module, etc.), which is a fairly common setup. Now, the important part of the question. I have a group of ARM 32-bit devices (ZedBoard w/ Xilinx Zynq SoC) that I would like to use in a cluster to submit jobs to.
I currently have PBS installed on my x86 host, and I an in the process of building PBS from source on one of the ARM 32-bit systems. This will probably take an hour or two to build since it’s a 2-core CPU in the SoC.
I just finished running a cross-compilation on my x86 machine of PBS, using the
--host=arm-linux-eabihf argument to the
./configure script, but I’m not sure how to “cleanly” transfer the extra config files like
/etc/profile.d/pbs.sh, etc. Mostly because I don’t know all of the files I’d need to transfer. I can probably figure out the permissions after the fact.
So, does this sound possible? If so, I may just wait for it to compile on the execution host (32-bit device) and see what happens. I’m compiling the same version of PBS on both machines.
Another question I have is that in the Big Book in section 22.214.171.124, it says the following:
Users who will submit jobs must have accounts at the server and at each execution host.
Does this mean every user has to have a Linux account on the physical execution hosts, or am I interpreting this incorrectly?
If this setup is possible, we’re going to have students submitting jobs to the boards, so we were planning on creating user account for them on the server, but I’m not sure how easy it would be to create them on all of the execution hosts as well, since we use PAM with Active Directory groups for the students’ user accounts.