TODO for prototype

• Watchdog
• Deployment
• Split NodeAgent into Nodecontroller and NodeAgent
  • NodeController - responsible for garbage collection of pods, collect health metrics
  • NodeAgent - responsible for pod lifecycle management
• Various deployment strategies
• Disruption budget
• Delta updates
• Concurrent updates
• StatefulSet - stable pod name
• More principled way to track state changes -- who owns what resources / fields / objects
• Proper state machine for the controller / nodeagent / allocator