raise SystemExit, 1 user = pwd.getpwuid(os.getuid())[0] jobspec = {'jobid':int(os.environ["COBALT_JOBID"]), 'user':user, 'true_mpi_args':arglist, 'walltime':j['walltime'], 'args':[], 'location':j['location'], 'outputdir':j['outputdir'], } for key in io_redirect: if io_redirect[key]: jobspec.update({key: io_redirect[key]}) try: scriptm = ComponentProxy("script-manager", defer=False) system = ComponentProxy("system", defer=False) # try adding job to queue_manager pgid = int(scriptm.invoke_mpi_from_script(jobspec)) # give the process a chance to get started before we check for it start = time.time() while True: r = system.get_process_groups([{'id':pgid, 'state':'*'}]) if r: break # we'll give it 90 seconds to get started if time.time() - start > 90: break time.sleep(5) while True: