There seems to be a 2 minutes
hard coded waiting time after a
parallel job has finished, which
is too long for my use case.
The place in the source code where
I suspect the problem
.../sge-8.0.0a/source/daemons/execd/exec_job.c
Lines 1294/1295
***************************************************************
/* JG: TODO: should at least be a define, better configurable */
fprintf(fp, "script_timeout=120\n");
***************************************************************
the actual waiting is in the
'do_pe_stop' method call,
there the method start_child is called
(Line 1172..1174 in shepherd):
*******************************************
if (timeout) {
shepherd_trace("using signal delivery delay of %d seconds",
timeout);
alarm(timeout);
************************************************
Is there any possibility to configure this
waiting time without recompiling the code?
Alain
_______________________________________________
users mailing list
[email protected]
https://gridengine.org/mailman/listinfo/users
