Hi,
I've given this some thought in the past... I'd like to point out that using condor_status to derive the list of execute nodes has one potential flaw: ifyou're really trying to hit *every* execute node, then nodes that aretemporarily not reporting in at the time (say, due to a crash) will not get ajob scheduled to run on them.
yes, without going the route of CFengine or similar, there is always the risk of loosing some of the machines, but so far I'm OK with that. For what it is worth, I ended up with some simple scripts in Perl to do it, which I put at http://angel-de-vicente.blogspot.com/2010/06/running-commandscript-in-all-machines.html
Cheers, Ángel de Vicente -- +---------------------------------------------+ | | | http://www.iac.es/galeria/angelv/ | | | | High Performance Computing Support PostDoc | | Instituto de Astrofísica de Canarias | | | +---------------------------------------------+