[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [Condor-users] Processor-architecture
- Date: Tue, 29 Jun 2004 11:17:59 +0100
- From: "Kewley, J (John)" <J.Kewley@xxxxxxxx>
- Subject: RE: [Condor-users] Processor-architecture
> I was just wondering, is there a way to give each client-machine local
> attributes via the condor_config or the condor_config.local?
> I want to expand the "arch"-info, because at our site it is
> not enough to
> know that the processor is an INTEL, when you need an INTEL
> P4 because of
> Any ideas?
You could set up your own policy, by adding in a PROC_TYPE (or whatever)
in the configuration for each machine in your pool (or at least ones of
an ARCH of a problematic type).
Any jobs for which such switches are an issue, can then specify in their
REQUIREMENTS that (PROC_TYPE == "P4")
The job will then only run on machines with that PROC_TYPE specified.
You could also build different executables for different OS/arch/proc
in the same style as the examples in the documentation for OS/ARCH.
A similar technique can also be used to distinguish between different OS
REQUIREMENTS = ((OSTYPE == LINUX) && (LINUX_MAJOR == "RedHat") &&
(LINUX_MINOR == "7.3"))
or libc versions.
User-defined attributes can also be used to specify whether a given machine
a particular os/w version installed.
Both these last 2 examples will only work if all relevent machines have
definitions in their configuration file.