DC/OS can run many different kinds of workloads, which are composed of tasks. DC/OS tasks are Mesos tasks that have been scheduled by either a DC/OS built-in scheduler or a scheduler service running on DC/OS.
When the scheduler launches a task, it specifies a Mesos Executor, which then executes the task. In Mesos, the scheduler and its executor(s) are called a framework, but within the broader context of DC/OS we often use the terms “scheduler”, “executor”, and “task” explicitly.
Mesos includes built-in executors that are available to all schedulers, but schedulers can also use their own executors.
- Command Executor - execute shell commands or Docker containers
- Default Executor (Mesos 1.1) - execute a group of shell commands or Docker containers
For more on Mesos executors, see the Mesos Framework Development Guide
Because the task system is so generic, users generally do not create or interact with tasks directly. Instead, schedulers often provide higher level abstractions.
DC/OS has two built-in schedulers:
- The Marathon scheduler provides services (Apps and Pods), which run continuously and in parallel. For more on Marathon services, see the Deploying Services docs or the Marathon docs.
- The Metronome scheduler provides jobs, which run immediately or on a defined schedule. For more on Metronome jobs, see the Deploying Jobs docs.
User space schedulers
Example user space schedulers:
- The Kafka scheduler provides Kafka brokers, which run as lifecycle managed Kafka nodes.
- The Cassandra scheduler provides Cassandra nodes, which run as lifecycle managed Cassandra nodes.
- The Spark scheduler (dispatcher) provides Spark jobs, which are themselves schedulers for Spark tasks.
For a full list of installable schedulers (and other packages), see the Mesosphere Universe package list.