Elastic Search Custom Plugin

  1. You can't. A rolling restart would be the way to go, node after node.
  2. Are you talking about out of memory exceptions or about the circuit breaker? Even then this is hard to tell on a per-request base, because a request might work just fine when there is no load, but it is a different case when there are hundreds of requests per second.
  3. What should this limit limit on?
  4. Use the reindex API