wtf is marathon?

Marathon is a tool that keeps your applications running continuously across a cluster of servers. If you have a web service or background job that needs to stay online no matter what, Marathon starts it, monitors it, and automatically restarts it if the machine it lives on crashes. You can think of it as the ignition system for your data center: once you tell it to run an app, it makes sure that app keeps running indefinitely. Under the hood, Marathon sits on top of Apache Mesos, software that pools your servers' resources together so they act like a single large computer. Marathon uses this pool to launch and manage your applications. You interact with it through a web interface or a REST API (a way for programs to talk to each other over the web) to start, stop, or scale apps up and down. If you run multiple copies of Marathon itself, they coordinate so only one acts as the leader at a time, and if that leader fails, another takes over automatically. This project is aimed at teams managing infrastructure at scale, companies like Airbnb, eBay, and PayPal are listed as users. A concrete use case would be a company running dozens of microservices across hundreds of servers. Instead of manually ensuring each service stays up, the team uses Marathon to enforce rules like "only run one instance per server" and to automatically route traffic to healthy instances using a load balancer. It also supports launching packaged applications via Docker containers. One notable aspect of Marathon is that it's a "meta framework", it can manage other Mesos frameworks, not just individual apps. This means you could use Marathon to ensure other critical services stay online after machine failures. The README notes it can launch anything that runs in a standard shell, giving it flexibility beyond just containerized workloads. It supports health checks, event notifications for external integrations, and basic authentication, making it a fairly complete management layer for long-running services.