Ficus kafka integration yet unresolved questions

[aerooneqq]

• Resources:
  • Drop reader metadata in Ficus after there are no updates from the application for some amount of time (contreversial), drop reader metadata if the target application has finished its execution
  • Persistent storage for reader metadata for long-running apps, now everything is in memory, for now it is ok, but later it will cause problems
  • Persistent storage for pipeline parts context values: again, now everything is in RAM, should be ok for now, but latter must store it somewhere