Currently we can drain already executing jobs to other machines using condor_drain instantaneously (i.e. need not wait for existing executing jobs to be completed) to other machines and then proceed for shutting down of the host.
As far as I know, condor_drain does not move jobs to other machines; they get evicted and rescheduled.
Also currently when we use "condor_off -schedd <SCHEDD_NAME>" which stops scheduler from receiving any new jobs and then wait for already submitted jobs, But we want something like draining the scheduler and moving to another machine instantaneously.
If you want the schedd to shut off immediately, you can add the -fast flag to condor_off. However, we don't support moving jobs from one schedd to another. We generally recommend that schedds are permanent, unless they're dedicate to the single user who controls if they're up or down.
- ToddM