Ignore entries in downtime when considering glideins matches
When an entry is in downtime it is still considered for the distribution of the jobs.
This causes fewer glideins to be requested to the entries that are up.
The job selection and assignment is considering the requests, not the currently running jobs and where they run.
To have a correct accounting of the entries in downtime (can still run glideins that were requested earlier, will not receive new glideins)
Considering both information at the same time would be a major rewrite of the whole logic of the Frontend.
Without running job information two approximate options are possible:
- keep considering the entries in downtime and underprovision - because no new glideins will start (current behavior)
- ignore the entries in downtime and overprovision - because some glideins may be running there (proposed)
The solution proposed is to add a switch that will allow in the Frontend (globally and in the groups) to select one of the 2 behaviors described above