Currently we only queue the processing jobs (via SSH login semaphores) themselves on the Controller, but not the rsync calls. To optimally pipeline the data transfers (to and from the Controller) in parallel to computation, we circumvent the queueing/locking mechanism (by using a different login other than ocrd).
But this means that too many jobs stacking on the queue can overflow the data volume on the Controller.
Currently we only queue the processing jobs (via SSH login semaphores) themselves on the Controller, but not the rsync calls. To optimally pipeline the data transfers (to and from the Controller) in parallel to computation, we circumvent the queueing/locking mechanism (by using a different login other than
ocrd).But this means that too many jobs stacking on the queue can overflow the data volume on the Controller.