that did the trick. I inserted the ulimit into the daemon startup script and restarted the service.
Thanks a lot for the hint, William. With kind regards, ulrich On 06/14/2016 09:48 AM, William Hay wrote: > On Mon, Jun 13, 2016 at 07:15:07PM +0200, Ulrich Hiller wrote: >> Dear William and Bill, >> >> thanks a lot for your answers. >> >> I already configured limits.conf a few days ago on all nodes. 'ulimit >> -n' (open files) gives 94000. That should be more than enough. > > How did you run that command? Have you tried submitting a job script > with that command in it to see what you get? > >> I do not understand where the breakpoint 252/253 comes from, and why it >> works with mpiexec directly on the node. Did i oversee a config issue? >> I am not totally convinced that it is not a gridengine issue. >> > > Did execd get started from a shell that was started before you modified > limits.conf? Changing limits.conf won't alter the limits for a running > execd or any process it spawns (barring cases where the shepherd spawns > an sshd or similar which changes limits via pam_limits. > > Assuming you are using an init script rather than a systemd unit file > to start the daemon you could add explicit invocations to ulimit into > /etc/sysconfig/sgeexecd (or other platform dependent file probably > something in /etc/defaults for debian derivatives) > set the various limits appropriately for the daemon if it doesn't > otherwise pick them up. No doubt systemd has some similar mechanism > for unit files. > > William > _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
