Per my knowledge, with Marathon/Chronos framework, applications can be deployed to slave nodes that fulfill their resource requirement. But I can't find any information on if those applications fail for internal defects, then if application developer or support engineer can logon the node to do diagnostic work? Collecting log file etc.
Regarding cloud native apps, logging service should be used to record what happened instead of local log file. I just read Oreilly book Migrating to Cloud Native application Architecture, quoted a paragraph as below:
Logs Rather than managing logfiles, treat logs as event streams, allowing the execution environment to collect, aggregate, index, and analyze the events via centralized services.