Disable Feature "Queuing Directly to a GPU"


I understand why this feature is here but when you have a big instance where everyone can use it or when you have “advanced users” that runs 8 jobs on 4 GPUs then it would be really really nice to have the possibility to disable this feature and allow only to use the scheduler.

Otherwise you never know why a GPU crashed now.


Thanks @biocit for the suggestion to allow disabling direct GPU queueing. We took a note of your suggestion.
Currently available options are:

  1. Replacing node-type scheduler lanes with cluster-type scheduler lanes. Because node GPU resource information would be stored in the cluster configuration instead of the CryoSPARC database, the Queuing Directly to a GPU option would no longer be available after all node-type targets or lanes have been removed with suitable remove_scheduler_target_node() and/or remove_scheduler_lane() cli commands.
  2. Alternatively, you could ask users to not use the Run on specific GPU option, perhaps in a CryoSPARC Message of the Day.

Thanks for the suggestions.

Cluster-type seams to be the way to go at the moment. But that seams quite complicated. A simple tick in the web interface would be easier :smiley: