When app scaling action is made more than zk_max_version times and then Marathon is restarted application configuration (and health checks conf) are lost.
There is WARN in log:
> WARN Cannot find health check configuration for [appname] and version [date], using most recent one. (mesosphere.marathon.core.health.impl.MarathonHealthCheckManager:ForkJoinPool-2-worker-151)
but as in: https://github.com/mesosphere/marathon/blob/17d27b6acf3e202f5113a401ab39e176135efb46/src/main/scala/mesosphere/marathon/core/health/impl/MarathonHealthCheckManager.scala#L172 health check configuration is not copied and task is in Unknown state.
- Create an application with health-check
- Scale it many times adding one instance
- Restart marathon
- The oldest instance will lost its health-check