The orphaned processes howto http://arc.liv.ac.uk/SGE/howto/remove_orphaned_processes.html now documents a couple of other techniques that may be useful at sites which have the same problems with orphans on Linux-based hosts as we do.
The proc-police daemon has been running for a while here in production here, where all jobs are meant to be tightly integrated. Note the hacked version, necessary on RH5, at least, and if anyone knows what might have broken the packet filtering, Brian Bockelman would be interested. The more adventurous could try the cpuset containment in the latest SGE snapshot <http://arc.liv.ac.uk/downloads/SGE/snapshots/>. The old cpuset implementation, as present e.g. in Red Hat 5, is emulated (or supposed to be) with cgroups in more recent versions of Linux, and its configuration and implementation is rather variable in current distributions. Feedback welcome, obviously. -- Community Grid Engine: http://arc.liv.ac.uk/SGE/ _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
