Rayson Ho <[email protected]> writes: > To make you even more confused, "-sync" is not really a job parameter, > but a qsub parameter. What I mean is that "-sync y" just controls the > way how qsub connects to the qmaster & waits for job status, but it > does not specify the scheduling requirements of a job. So logically > JSV does not cover -sync.
Is JSV defined somewhere to deal with scheduling, rather than submission? It seems logical to have -sync available to a submission verifier. > Lastly, if you are concerned about performance, read this DanT's blog entry: > > http://blogs.oracle.com/templedf/entry/performance_considerations_for_jsv_scripts I'm unconvinced in the absence of more realistic measurements with bash or similar. I think http://arc.liv.ac.uk/repos/darcs/sge/source/dist/util/resources/jsv/jsv_include.sh should normally only spawn 3 external commands for each onvocation under bash -- 2 `date's, and 1 GE `arch'. It doesn't need read_raw and write_raw in particular. Changes welcome for other shells. On the other hand, security in server-side sh jsvs (running privileged) would bother me. -- Excuse the typping -- I have a broken wrist _______________________________________________ users mailing list [email protected] https://gridengine.org/mailman/listinfo/users
