Previously most workloads on the Kubernetes cluster have been web services and other continuous jobs where a restart and a move to another node would not have mattered. This assumption changes when the jobs framework introduces cron jobs. This task is to:

• check if running jobs will not misbehave when they are restarted
  • TODO: should jobs reschedule or not if they don't complete? especially relevant for one-off jobs
• consider adding some delay to let running jobs complete when nodes are being drained