File staging and pilotjob scheduling are concurrent - i.e., file staging might happen while the pilotjob has been already scheduled on the remote resource. If the pilotjob becomes active before the file staging has completed, CUs wait for the file staging to be completed before being executed (this is controlled in sagapilot right?). The implicit and fallacious assumption is that enough walltime will always be available for the pilotjob to wait for the file staging to complete and then for the CUs to execute as planned.
TODO:
Integrate in the calculation of the pilotjob walltime an educated guess of the time needed for file stage-in.
File staging and pilotjob scheduling are concurrent - i.e., file staging might happen while the pilotjob has been already scheduled on the remote resource. If the pilotjob becomes active before the file staging has completed, CUs wait for the file staging to be completed before being executed (this is controlled in sagapilot right?). The implicit and fallacious assumption is that enough walltime will always be available for the pilotjob to wait for the file staging to complete and then for the CUs to execute as planned.
TODO: