maintenance mode 3.3 notes

• Support for vmware 5.0 and 5.1
• There is a clear overhead to setting up the environment for vmware setup
• Are there conditions under which evacuation can fail?
  • What is the behaviour in the resource exhaustion case?
    • Defragging? Over-committing?
• What are the expected performance characteristics?
• Planned hardware migration is the fundamental use case
• Failure feedback:
  • Immediate -- it is possible to, i will reserve resources for, and then perform migration
  • Batched/Eventual -- keep trying to do migration until it succeeds
• Administrator can quarantine resources for future when current usage levels prevent immediate evacuation
• Choosing the destination node is not a part of the MVP for 3.3
• Vast majority not using shared storage
  • Trending towards commodity storage
• Best practice? Reserved resources for the purposes of supporting migration
  • Would require support for marking resources as reserved
• Three scenarios
  • Resources available
  • Resource over-committing needed
    • Additional over-commiting factors to support maintenance emergencies?
  • Resources not available
• Least common denominator masking for CPU types
• What do we have to do wrt w/ concurrency
  • Handle only one operation at a time in 3.3
• Support for multiple network interfaces