Closed FinchPowers closed 8 years ago
Not too good after all. I looked at our production cluster and the job duration average is currently 8 seconds. It means that to add 2 nodes we would need to have at least 451 jobs in the queue and that our current slot usage would be maxed.
As this is low value, I will drop it entirely for now.
If the cluster has a running average of 15 minutes and sees 10 jobs waiting, will no longer add 10 nodes but 3 instead.